1 // SPDX-License-Identifier: GPL-2.0 2 /* 3 * linux/fs/ext4/super.c 4 * 5 * Copyright (C) 1992, 1993, 1994, 1995 6 * Remy Card (card@masi.ibp.fr) 7 * Laboratoire MASI - Institut Blaise Pascal 8 * Universite Pierre et Marie Curie (Paris VI) 9 * 10 * from 11 * 12 * linux/fs/minix/inode.c 13 * 14 * Copyright (C) 1991, 1992 Linus Torvalds 15 * 16 * Big-endian to little-endian byte-swapping/bitmaps by 17 * David S. Miller (davem@caip.rutgers.edu), 1995 18 */ 19 20 #include <linux/module.h> 21 #include <linux/string.h> 22 #include <linux/fs.h> 23 #include <linux/time.h> 24 #include <linux/vmalloc.h> 25 #include <linux/slab.h> 26 #include <linux/init.h> 27 #include <linux/blkdev.h> 28 #include <linux/backing-dev.h> 29 #include <linux/parser.h> 30 #include <linux/buffer_head.h> 31 #include <linux/exportfs.h> 32 #include <linux/vfs.h> 33 #include <linux/random.h> 34 #include <linux/mount.h> 35 #include <linux/namei.h> 36 #include <linux/quotaops.h> 37 #include <linux/seq_file.h> 38 #include <linux/ctype.h> 39 #include <linux/log2.h> 40 #include <linux/crc16.h> 41 #include <linux/dax.h> 42 #include <linux/cleancache.h> 43 #include <linux/uaccess.h> 44 #include <linux/iversion.h> 45 #include <linux/unicode.h> 46 47 #include <linux/kthread.h> 48 #include <linux/freezer.h> 49 50 #include "ext4.h" 51 #include "ext4_extents.h" /* Needed for trace points definition */ 52 #include "ext4_jbd2.h" 53 #include "xattr.h" 54 #include "acl.h" 55 #include "mballoc.h" 56 #include "fsmap.h" 57 58 #define CREATE_TRACE_POINTS 59 #include <trace/events/ext4.h> 60 61 static struct ext4_lazy_init *ext4_li_info; 62 static struct mutex ext4_li_mtx; 63 static struct ratelimit_state ext4_mount_msg_ratelimit; 64 65 static int ext4_load_journal(struct super_block *, struct ext4_super_block *, 66 unsigned long journal_devnum); 67 static int ext4_show_options(struct seq_file *seq, struct dentry *root); 68 static int ext4_commit_super(struct super_block *sb, int sync); 69 static void ext4_mark_recovery_complete(struct super_block *sb, 70 struct ext4_super_block *es); 71 static void ext4_clear_journal_err(struct super_block *sb, 72 struct ext4_super_block *es); 73 static int ext4_sync_fs(struct super_block *sb, int wait); 74 static int ext4_remount(struct super_block *sb, int *flags, char *data); 75 static int ext4_statfs(struct dentry *dentry, struct kstatfs *buf); 76 static int ext4_unfreeze(struct super_block *sb); 77 static int ext4_freeze(struct super_block *sb); 78 static struct dentry *ext4_mount(struct file_system_type *fs_type, int flags, 79 const char *dev_name, void *data); 80 static inline int ext2_feature_set_ok(struct super_block *sb); 81 static inline int ext3_feature_set_ok(struct super_block *sb); 82 static int ext4_feature_set_ok(struct super_block *sb, int readonly); 83 static void ext4_destroy_lazyinit_thread(void); 84 static void ext4_unregister_li_request(struct super_block *sb); 85 static void ext4_clear_request_list(void); 86 static struct inode *ext4_get_journal_inode(struct super_block *sb, 87 unsigned int journal_inum); 88 89 /* 90 * Lock ordering 91 * 92 * Note the difference between i_mmap_sem (EXT4_I(inode)->i_mmap_sem) and 93 * i_mmap_rwsem (inode->i_mmap_rwsem)! 94 * 95 * page fault path: 96 * mmap_sem -> sb_start_pagefault -> i_mmap_sem (r) -> transaction start -> 97 * page lock -> i_data_sem (rw) 98 * 99 * buffered write path: 100 * sb_start_write -> i_mutex -> mmap_sem 101 * sb_start_write -> i_mutex -> transaction start -> page lock -> 102 * i_data_sem (rw) 103 * 104 * truncate: 105 * sb_start_write -> i_mutex -> i_mmap_sem (w) -> i_mmap_rwsem (w) -> page lock 106 * sb_start_write -> i_mutex -> i_mmap_sem (w) -> transaction start -> 107 * i_data_sem (rw) 108 * 109 * direct IO: 110 * sb_start_write -> i_mutex -> mmap_sem 111 * sb_start_write -> i_mutex -> transaction start -> i_data_sem (rw) 112 * 113 * writepages: 114 * transaction start -> page lock(s) -> i_data_sem (rw) 115 */ 116 117 #if !defined(CONFIG_EXT2_FS) && !defined(CONFIG_EXT2_FS_MODULE) && defined(CONFIG_EXT4_USE_FOR_EXT2) 118 static struct file_system_type ext2_fs_type = { 119 .owner = THIS_MODULE, 120 .name = "ext2", 121 .mount = ext4_mount, 122 .kill_sb = kill_block_super, 123 .fs_flags = FS_REQUIRES_DEV, 124 }; 125 MODULE_ALIAS_FS("ext2"); 126 MODULE_ALIAS("ext2"); 127 #define IS_EXT2_SB(sb) ((sb)->s_bdev->bd_holder == &ext2_fs_type) 128 #else 129 #define IS_EXT2_SB(sb) (0) 130 #endif 131 132 133 static struct file_system_type ext3_fs_type = { 134 .owner = THIS_MODULE, 135 .name = "ext3", 136 .mount = ext4_mount, 137 .kill_sb = kill_block_super, 138 .fs_flags = FS_REQUIRES_DEV, 139 }; 140 MODULE_ALIAS_FS("ext3"); 141 MODULE_ALIAS("ext3"); 142 #define IS_EXT3_SB(sb) ((sb)->s_bdev->bd_holder == &ext3_fs_type) 143 144 /* 145 * This works like sb_bread() except it uses ERR_PTR for error 146 * returns. Currently with sb_bread it's impossible to distinguish 147 * between ENOMEM and EIO situations (since both result in a NULL 148 * return. 149 */ 150 struct buffer_head * 151 ext4_sb_bread(struct super_block *sb, sector_t block, int op_flags) 152 { 153 struct buffer_head *bh = sb_getblk(sb, block); 154 155 if (bh == NULL) 156 return ERR_PTR(-ENOMEM); 157 if (buffer_uptodate(bh)) 158 return bh; 159 ll_rw_block(REQ_OP_READ, REQ_META | op_flags, 1, &bh); 160 wait_on_buffer(bh); 161 if (buffer_uptodate(bh)) 162 return bh; 163 put_bh(bh); 164 return ERR_PTR(-EIO); 165 } 166 167 static int ext4_verify_csum_type(struct super_block *sb, 168 struct ext4_super_block *es) 169 { 170 if (!ext4_has_feature_metadata_csum(sb)) 171 return 1; 172 173 return es->s_checksum_type == EXT4_CRC32C_CHKSUM; 174 } 175 176 static __le32 ext4_superblock_csum(struct super_block *sb, 177 struct ext4_super_block *es) 178 { 179 struct ext4_sb_info *sbi = EXT4_SB(sb); 180 int offset = offsetof(struct ext4_super_block, s_checksum); 181 __u32 csum; 182 183 csum = ext4_chksum(sbi, ~0, (char *)es, offset); 184 185 return cpu_to_le32(csum); 186 } 187 188 static int ext4_superblock_csum_verify(struct super_block *sb, 189 struct ext4_super_block *es) 190 { 191 if (!ext4_has_metadata_csum(sb)) 192 return 1; 193 194 return es->s_checksum == ext4_superblock_csum(sb, es); 195 } 196 197 void ext4_superblock_csum_set(struct super_block *sb) 198 { 199 struct ext4_super_block *es = EXT4_SB(sb)->s_es; 200 201 if (!ext4_has_metadata_csum(sb)) 202 return; 203 204 es->s_checksum = ext4_superblock_csum(sb, es); 205 } 206 207 void *ext4_kvmalloc(size_t size, gfp_t flags) 208 { 209 void *ret; 210 211 ret = kmalloc(size, flags | __GFP_NOWARN); 212 if (!ret) 213 ret = __vmalloc(size, flags, PAGE_KERNEL); 214 return ret; 215 } 216 217 void *ext4_kvzalloc(size_t size, gfp_t flags) 218 { 219 void *ret; 220 221 ret = kzalloc(size, flags | __GFP_NOWARN); 222 if (!ret) 223 ret = __vmalloc(size, flags | __GFP_ZERO, PAGE_KERNEL); 224 return ret; 225 } 226 227 ext4_fsblk_t ext4_block_bitmap(struct super_block *sb, 228 struct ext4_group_desc *bg) 229 { 230 return le32_to_cpu(bg->bg_block_bitmap_lo) | 231 (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT ? 232 (ext4_fsblk_t)le32_to_cpu(bg->bg_block_bitmap_hi) << 32 : 0); 233 } 234 235 ext4_fsblk_t ext4_inode_bitmap(struct super_block *sb, 236 struct ext4_group_desc *bg) 237 { 238 return le32_to_cpu(bg->bg_inode_bitmap_lo) | 239 (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT ? 240 (ext4_fsblk_t)le32_to_cpu(bg->bg_inode_bitmap_hi) << 32 : 0); 241 } 242 243 ext4_fsblk_t ext4_inode_table(struct super_block *sb, 244 struct ext4_group_desc *bg) 245 { 246 return le32_to_cpu(bg->bg_inode_table_lo) | 247 (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT ? 248 (ext4_fsblk_t)le32_to_cpu(bg->bg_inode_table_hi) << 32 : 0); 249 } 250 251 __u32 ext4_free_group_clusters(struct super_block *sb, 252 struct ext4_group_desc *bg) 253 { 254 return le16_to_cpu(bg->bg_free_blocks_count_lo) | 255 (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT ? 256 (__u32)le16_to_cpu(bg->bg_free_blocks_count_hi) << 16 : 0); 257 } 258 259 __u32 ext4_free_inodes_count(struct super_block *sb, 260 struct ext4_group_desc *bg) 261 { 262 return le16_to_cpu(bg->bg_free_inodes_count_lo) | 263 (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT ? 264 (__u32)le16_to_cpu(bg->bg_free_inodes_count_hi) << 16 : 0); 265 } 266 267 __u32 ext4_used_dirs_count(struct super_block *sb, 268 struct ext4_group_desc *bg) 269 { 270 return le16_to_cpu(bg->bg_used_dirs_count_lo) | 271 (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT ? 272 (__u32)le16_to_cpu(bg->bg_used_dirs_count_hi) << 16 : 0); 273 } 274 275 __u32 ext4_itable_unused_count(struct super_block *sb, 276 struct ext4_group_desc *bg) 277 { 278 return le16_to_cpu(bg->bg_itable_unused_lo) | 279 (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT ? 280 (__u32)le16_to_cpu(bg->bg_itable_unused_hi) << 16 : 0); 281 } 282 283 void ext4_block_bitmap_set(struct super_block *sb, 284 struct ext4_group_desc *bg, ext4_fsblk_t blk) 285 { 286 bg->bg_block_bitmap_lo = cpu_to_le32((u32)blk); 287 if (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT) 288 bg->bg_block_bitmap_hi = cpu_to_le32(blk >> 32); 289 } 290 291 void ext4_inode_bitmap_set(struct super_block *sb, 292 struct ext4_group_desc *bg, ext4_fsblk_t blk) 293 { 294 bg->bg_inode_bitmap_lo = cpu_to_le32((u32)blk); 295 if (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT) 296 bg->bg_inode_bitmap_hi = cpu_to_le32(blk >> 32); 297 } 298 299 void ext4_inode_table_set(struct super_block *sb, 300 struct ext4_group_desc *bg, ext4_fsblk_t blk) 301 { 302 bg->bg_inode_table_lo = cpu_to_le32((u32)blk); 303 if (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT) 304 bg->bg_inode_table_hi = cpu_to_le32(blk >> 32); 305 } 306 307 void ext4_free_group_clusters_set(struct super_block *sb, 308 struct ext4_group_desc *bg, __u32 count) 309 { 310 bg->bg_free_blocks_count_lo = cpu_to_le16((__u16)count); 311 if (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT) 312 bg->bg_free_blocks_count_hi = cpu_to_le16(count >> 16); 313 } 314 315 void ext4_free_inodes_set(struct super_block *sb, 316 struct ext4_group_desc *bg, __u32 count) 317 { 318 bg->bg_free_inodes_count_lo = cpu_to_le16((__u16)count); 319 if (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT) 320 bg->bg_free_inodes_count_hi = cpu_to_le16(count >> 16); 321 } 322 323 void ext4_used_dirs_set(struct super_block *sb, 324 struct ext4_group_desc *bg, __u32 count) 325 { 326 bg->bg_used_dirs_count_lo = cpu_to_le16((__u16)count); 327 if (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT) 328 bg->bg_used_dirs_count_hi = cpu_to_le16(count >> 16); 329 } 330 331 void ext4_itable_unused_set(struct super_block *sb, 332 struct ext4_group_desc *bg, __u32 count) 333 { 334 bg->bg_itable_unused_lo = cpu_to_le16((__u16)count); 335 if (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT) 336 bg->bg_itable_unused_hi = cpu_to_le16(count >> 16); 337 } 338 339 static void __ext4_update_tstamp(__le32 *lo, __u8 *hi) 340 { 341 time64_t now = ktime_get_real_seconds(); 342 343 now = clamp_val(now, 0, (1ull << 40) - 1); 344 345 *lo = cpu_to_le32(lower_32_bits(now)); 346 *hi = upper_32_bits(now); 347 } 348 349 static time64_t __ext4_get_tstamp(__le32 *lo, __u8 *hi) 350 { 351 return ((time64_t)(*hi) << 32) + le32_to_cpu(*lo); 352 } 353 #define ext4_update_tstamp(es, tstamp) \ 354 __ext4_update_tstamp(&(es)->tstamp, &(es)->tstamp ## _hi) 355 #define ext4_get_tstamp(es, tstamp) \ 356 __ext4_get_tstamp(&(es)->tstamp, &(es)->tstamp ## _hi) 357 358 static void __save_error_info(struct super_block *sb, const char *func, 359 unsigned int line) 360 { 361 struct ext4_super_block *es = EXT4_SB(sb)->s_es; 362 363 EXT4_SB(sb)->s_mount_state |= EXT4_ERROR_FS; 364 if (bdev_read_only(sb->s_bdev)) 365 return; 366 es->s_state |= cpu_to_le16(EXT4_ERROR_FS); 367 ext4_update_tstamp(es, s_last_error_time); 368 strncpy(es->s_last_error_func, func, sizeof(es->s_last_error_func)); 369 es->s_last_error_line = cpu_to_le32(line); 370 if (!es->s_first_error_time) { 371 es->s_first_error_time = es->s_last_error_time; 372 es->s_first_error_time_hi = es->s_last_error_time_hi; 373 strncpy(es->s_first_error_func, func, 374 sizeof(es->s_first_error_func)); 375 es->s_first_error_line = cpu_to_le32(line); 376 es->s_first_error_ino = es->s_last_error_ino; 377 es->s_first_error_block = es->s_last_error_block; 378 } 379 /* 380 * Start the daily error reporting function if it hasn't been 381 * started already 382 */ 383 if (!es->s_error_count) 384 mod_timer(&EXT4_SB(sb)->s_err_report, jiffies + 24*60*60*HZ); 385 le32_add_cpu(&es->s_error_count, 1); 386 } 387 388 static void save_error_info(struct super_block *sb, const char *func, 389 unsigned int line) 390 { 391 __save_error_info(sb, func, line); 392 ext4_commit_super(sb, 1); 393 } 394 395 /* 396 * The del_gendisk() function uninitializes the disk-specific data 397 * structures, including the bdi structure, without telling anyone 398 * else. Once this happens, any attempt to call mark_buffer_dirty() 399 * (for example, by ext4_commit_super), will cause a kernel OOPS. 400 * This is a kludge to prevent these oops until we can put in a proper 401 * hook in del_gendisk() to inform the VFS and file system layers. 402 */ 403 static int block_device_ejected(struct super_block *sb) 404 { 405 struct inode *bd_inode = sb->s_bdev->bd_inode; 406 struct backing_dev_info *bdi = inode_to_bdi(bd_inode); 407 408 return bdi->dev == NULL; 409 } 410 411 static void ext4_journal_commit_callback(journal_t *journal, transaction_t *txn) 412 { 413 struct super_block *sb = journal->j_private; 414 struct ext4_sb_info *sbi = EXT4_SB(sb); 415 int error = is_journal_aborted(journal); 416 struct ext4_journal_cb_entry *jce; 417 418 BUG_ON(txn->t_state == T_FINISHED); 419 420 ext4_process_freed_data(sb, txn->t_tid); 421 422 spin_lock(&sbi->s_md_lock); 423 while (!list_empty(&txn->t_private_list)) { 424 jce = list_entry(txn->t_private_list.next, 425 struct ext4_journal_cb_entry, jce_list); 426 list_del_init(&jce->jce_list); 427 spin_unlock(&sbi->s_md_lock); 428 jce->jce_func(sb, jce, error); 429 spin_lock(&sbi->s_md_lock); 430 } 431 spin_unlock(&sbi->s_md_lock); 432 } 433 434 static bool system_going_down(void) 435 { 436 return system_state == SYSTEM_HALT || system_state == SYSTEM_POWER_OFF 437 || system_state == SYSTEM_RESTART; 438 } 439 440 /* Deal with the reporting of failure conditions on a filesystem such as 441 * inconsistencies detected or read IO failures. 442 * 443 * On ext2, we can store the error state of the filesystem in the 444 * superblock. That is not possible on ext4, because we may have other 445 * write ordering constraints on the superblock which prevent us from 446 * writing it out straight away; and given that the journal is about to 447 * be aborted, we can't rely on the current, or future, transactions to 448 * write out the superblock safely. 449 * 450 * We'll just use the jbd2_journal_abort() error code to record an error in 451 * the journal instead. On recovery, the journal will complain about 452 * that error until we've noted it down and cleared it. 453 */ 454 455 static void ext4_handle_error(struct super_block *sb) 456 { 457 if (test_opt(sb, WARN_ON_ERROR)) 458 WARN_ON_ONCE(1); 459 460 if (sb_rdonly(sb)) 461 return; 462 463 if (!test_opt(sb, ERRORS_CONT)) { 464 journal_t *journal = EXT4_SB(sb)->s_journal; 465 466 EXT4_SB(sb)->s_mount_flags |= EXT4_MF_FS_ABORTED; 467 if (journal) 468 jbd2_journal_abort(journal, -EIO); 469 } 470 /* 471 * We force ERRORS_RO behavior when system is rebooting. Otherwise we 472 * could panic during 'reboot -f' as the underlying device got already 473 * disabled. 474 */ 475 if (test_opt(sb, ERRORS_RO) || system_going_down()) { 476 ext4_msg(sb, KERN_CRIT, "Remounting filesystem read-only"); 477 /* 478 * Make sure updated value of ->s_mount_flags will be visible 479 * before ->s_flags update 480 */ 481 smp_wmb(); 482 sb->s_flags |= SB_RDONLY; 483 } else if (test_opt(sb, ERRORS_PANIC)) { 484 if (EXT4_SB(sb)->s_journal && 485 !(EXT4_SB(sb)->s_journal->j_flags & JBD2_REC_ERR)) 486 return; 487 panic("EXT4-fs (device %s): panic forced after error\n", 488 sb->s_id); 489 } 490 } 491 492 #define ext4_error_ratelimit(sb) \ 493 ___ratelimit(&(EXT4_SB(sb)->s_err_ratelimit_state), \ 494 "EXT4-fs error") 495 496 void __ext4_error(struct super_block *sb, const char *function, 497 unsigned int line, const char *fmt, ...) 498 { 499 struct va_format vaf; 500 va_list args; 501 502 if (unlikely(ext4_forced_shutdown(EXT4_SB(sb)))) 503 return; 504 505 trace_ext4_error(sb, function, line); 506 if (ext4_error_ratelimit(sb)) { 507 va_start(args, fmt); 508 vaf.fmt = fmt; 509 vaf.va = &args; 510 printk(KERN_CRIT 511 "EXT4-fs error (device %s): %s:%d: comm %s: %pV\n", 512 sb->s_id, function, line, current->comm, &vaf); 513 va_end(args); 514 } 515 save_error_info(sb, function, line); 516 ext4_handle_error(sb); 517 } 518 519 void __ext4_error_inode(struct inode *inode, const char *function, 520 unsigned int line, ext4_fsblk_t block, 521 const char *fmt, ...) 522 { 523 va_list args; 524 struct va_format vaf; 525 struct ext4_super_block *es = EXT4_SB(inode->i_sb)->s_es; 526 527 if (unlikely(ext4_forced_shutdown(EXT4_SB(inode->i_sb)))) 528 return; 529 530 trace_ext4_error(inode->i_sb, function, line); 531 es->s_last_error_ino = cpu_to_le32(inode->i_ino); 532 es->s_last_error_block = cpu_to_le64(block); 533 if (ext4_error_ratelimit(inode->i_sb)) { 534 va_start(args, fmt); 535 vaf.fmt = fmt; 536 vaf.va = &args; 537 if (block) 538 printk(KERN_CRIT "EXT4-fs error (device %s): %s:%d: " 539 "inode #%lu: block %llu: comm %s: %pV\n", 540 inode->i_sb->s_id, function, line, inode->i_ino, 541 block, current->comm, &vaf); 542 else 543 printk(KERN_CRIT "EXT4-fs error (device %s): %s:%d: " 544 "inode #%lu: comm %s: %pV\n", 545 inode->i_sb->s_id, function, line, inode->i_ino, 546 current->comm, &vaf); 547 va_end(args); 548 } 549 save_error_info(inode->i_sb, function, line); 550 ext4_handle_error(inode->i_sb); 551 } 552 553 void __ext4_error_file(struct file *file, const char *function, 554 unsigned int line, ext4_fsblk_t block, 555 const char *fmt, ...) 556 { 557 va_list args; 558 struct va_format vaf; 559 struct ext4_super_block *es; 560 struct inode *inode = file_inode(file); 561 char pathname[80], *path; 562 563 if (unlikely(ext4_forced_shutdown(EXT4_SB(inode->i_sb)))) 564 return; 565 566 trace_ext4_error(inode->i_sb, function, line); 567 es = EXT4_SB(inode->i_sb)->s_es; 568 es->s_last_error_ino = cpu_to_le32(inode->i_ino); 569 if (ext4_error_ratelimit(inode->i_sb)) { 570 path = file_path(file, pathname, sizeof(pathname)); 571 if (IS_ERR(path)) 572 path = "(unknown)"; 573 va_start(args, fmt); 574 vaf.fmt = fmt; 575 vaf.va = &args; 576 if (block) 577 printk(KERN_CRIT 578 "EXT4-fs error (device %s): %s:%d: inode #%lu: " 579 "block %llu: comm %s: path %s: %pV\n", 580 inode->i_sb->s_id, function, line, inode->i_ino, 581 block, current->comm, path, &vaf); 582 else 583 printk(KERN_CRIT 584 "EXT4-fs error (device %s): %s:%d: inode #%lu: " 585 "comm %s: path %s: %pV\n", 586 inode->i_sb->s_id, function, line, inode->i_ino, 587 current->comm, path, &vaf); 588 va_end(args); 589 } 590 save_error_info(inode->i_sb, function, line); 591 ext4_handle_error(inode->i_sb); 592 } 593 594 const char *ext4_decode_error(struct super_block *sb, int errno, 595 char nbuf[16]) 596 { 597 char *errstr = NULL; 598 599 switch (errno) { 600 case -EFSCORRUPTED: 601 errstr = "Corrupt filesystem"; 602 break; 603 case -EFSBADCRC: 604 errstr = "Filesystem failed CRC"; 605 break; 606 case -EIO: 607 errstr = "IO failure"; 608 break; 609 case -ENOMEM: 610 errstr = "Out of memory"; 611 break; 612 case -EROFS: 613 if (!sb || (EXT4_SB(sb)->s_journal && 614 EXT4_SB(sb)->s_journal->j_flags & JBD2_ABORT)) 615 errstr = "Journal has aborted"; 616 else 617 errstr = "Readonly filesystem"; 618 break; 619 default: 620 /* If the caller passed in an extra buffer for unknown 621 * errors, textualise them now. Else we just return 622 * NULL. */ 623 if (nbuf) { 624 /* Check for truncated error codes... */ 625 if (snprintf(nbuf, 16, "error %d", -errno) >= 0) 626 errstr = nbuf; 627 } 628 break; 629 } 630 631 return errstr; 632 } 633 634 /* __ext4_std_error decodes expected errors from journaling functions 635 * automatically and invokes the appropriate error response. */ 636 637 void __ext4_std_error(struct super_block *sb, const char *function, 638 unsigned int line, int errno) 639 { 640 char nbuf[16]; 641 const char *errstr; 642 643 if (unlikely(ext4_forced_shutdown(EXT4_SB(sb)))) 644 return; 645 646 /* Special case: if the error is EROFS, and we're not already 647 * inside a transaction, then there's really no point in logging 648 * an error. */ 649 if (errno == -EROFS && journal_current_handle() == NULL && sb_rdonly(sb)) 650 return; 651 652 if (ext4_error_ratelimit(sb)) { 653 errstr = ext4_decode_error(sb, errno, nbuf); 654 printk(KERN_CRIT "EXT4-fs error (device %s) in %s:%d: %s\n", 655 sb->s_id, function, line, errstr); 656 } 657 658 save_error_info(sb, function, line); 659 ext4_handle_error(sb); 660 } 661 662 /* 663 * ext4_abort is a much stronger failure handler than ext4_error. The 664 * abort function may be used to deal with unrecoverable failures such 665 * as journal IO errors or ENOMEM at a critical moment in log management. 666 * 667 * We unconditionally force the filesystem into an ABORT|READONLY state, 668 * unless the error response on the fs has been set to panic in which 669 * case we take the easy way out and panic immediately. 670 */ 671 672 void __ext4_abort(struct super_block *sb, const char *function, 673 unsigned int line, const char *fmt, ...) 674 { 675 struct va_format vaf; 676 va_list args; 677 678 if (unlikely(ext4_forced_shutdown(EXT4_SB(sb)))) 679 return; 680 681 save_error_info(sb, function, line); 682 va_start(args, fmt); 683 vaf.fmt = fmt; 684 vaf.va = &args; 685 printk(KERN_CRIT "EXT4-fs error (device %s): %s:%d: %pV\n", 686 sb->s_id, function, line, &vaf); 687 va_end(args); 688 689 if (sb_rdonly(sb) == 0) { 690 ext4_msg(sb, KERN_CRIT, "Remounting filesystem read-only"); 691 EXT4_SB(sb)->s_mount_flags |= EXT4_MF_FS_ABORTED; 692 /* 693 * Make sure updated value of ->s_mount_flags will be visible 694 * before ->s_flags update 695 */ 696 smp_wmb(); 697 sb->s_flags |= SB_RDONLY; 698 if (EXT4_SB(sb)->s_journal) 699 jbd2_journal_abort(EXT4_SB(sb)->s_journal, -EIO); 700 save_error_info(sb, function, line); 701 } 702 if (test_opt(sb, ERRORS_PANIC) && !system_going_down()) { 703 if (EXT4_SB(sb)->s_journal && 704 !(EXT4_SB(sb)->s_journal->j_flags & JBD2_REC_ERR)) 705 return; 706 panic("EXT4-fs panic from previous error\n"); 707 } 708 } 709 710 void __ext4_msg(struct super_block *sb, 711 const char *prefix, const char *fmt, ...) 712 { 713 struct va_format vaf; 714 va_list args; 715 716 if (!___ratelimit(&(EXT4_SB(sb)->s_msg_ratelimit_state), "EXT4-fs")) 717 return; 718 719 va_start(args, fmt); 720 vaf.fmt = fmt; 721 vaf.va = &args; 722 printk("%sEXT4-fs (%s): %pV\n", prefix, sb->s_id, &vaf); 723 va_end(args); 724 } 725 726 #define ext4_warning_ratelimit(sb) \ 727 ___ratelimit(&(EXT4_SB(sb)->s_warning_ratelimit_state), \ 728 "EXT4-fs warning") 729 730 void __ext4_warning(struct super_block *sb, const char *function, 731 unsigned int line, const char *fmt, ...) 732 { 733 struct va_format vaf; 734 va_list args; 735 736 if (!ext4_warning_ratelimit(sb)) 737 return; 738 739 va_start(args, fmt); 740 vaf.fmt = fmt; 741 vaf.va = &args; 742 printk(KERN_WARNING "EXT4-fs warning (device %s): %s:%d: %pV\n", 743 sb->s_id, function, line, &vaf); 744 va_end(args); 745 } 746 747 void __ext4_warning_inode(const struct inode *inode, const char *function, 748 unsigned int line, const char *fmt, ...) 749 { 750 struct va_format vaf; 751 va_list args; 752 753 if (!ext4_warning_ratelimit(inode->i_sb)) 754 return; 755 756 va_start(args, fmt); 757 vaf.fmt = fmt; 758 vaf.va = &args; 759 printk(KERN_WARNING "EXT4-fs warning (device %s): %s:%d: " 760 "inode #%lu: comm %s: %pV\n", inode->i_sb->s_id, 761 function, line, inode->i_ino, current->comm, &vaf); 762 va_end(args); 763 } 764 765 void __ext4_grp_locked_error(const char *function, unsigned int line, 766 struct super_block *sb, ext4_group_t grp, 767 unsigned long ino, ext4_fsblk_t block, 768 const char *fmt, ...) 769 __releases(bitlock) 770 __acquires(bitlock) 771 { 772 struct va_format vaf; 773 va_list args; 774 struct ext4_super_block *es = EXT4_SB(sb)->s_es; 775 776 if (unlikely(ext4_forced_shutdown(EXT4_SB(sb)))) 777 return; 778 779 trace_ext4_error(sb, function, line); 780 es->s_last_error_ino = cpu_to_le32(ino); 781 es->s_last_error_block = cpu_to_le64(block); 782 __save_error_info(sb, function, line); 783 784 if (ext4_error_ratelimit(sb)) { 785 va_start(args, fmt); 786 vaf.fmt = fmt; 787 vaf.va = &args; 788 printk(KERN_CRIT "EXT4-fs error (device %s): %s:%d: group %u, ", 789 sb->s_id, function, line, grp); 790 if (ino) 791 printk(KERN_CONT "inode %lu: ", ino); 792 if (block) 793 printk(KERN_CONT "block %llu:", 794 (unsigned long long) block); 795 printk(KERN_CONT "%pV\n", &vaf); 796 va_end(args); 797 } 798 799 if (test_opt(sb, WARN_ON_ERROR)) 800 WARN_ON_ONCE(1); 801 802 if (test_opt(sb, ERRORS_CONT)) { 803 ext4_commit_super(sb, 0); 804 return; 805 } 806 807 ext4_unlock_group(sb, grp); 808 ext4_commit_super(sb, 1); 809 ext4_handle_error(sb); 810 /* 811 * We only get here in the ERRORS_RO case; relocking the group 812 * may be dangerous, but nothing bad will happen since the 813 * filesystem will have already been marked read/only and the 814 * journal has been aborted. We return 1 as a hint to callers 815 * who might what to use the return value from 816 * ext4_grp_locked_error() to distinguish between the 817 * ERRORS_CONT and ERRORS_RO case, and perhaps return more 818 * aggressively from the ext4 function in question, with a 819 * more appropriate error code. 820 */ 821 ext4_lock_group(sb, grp); 822 return; 823 } 824 825 void ext4_mark_group_bitmap_corrupted(struct super_block *sb, 826 ext4_group_t group, 827 unsigned int flags) 828 { 829 struct ext4_sb_info *sbi = EXT4_SB(sb); 830 struct ext4_group_info *grp = ext4_get_group_info(sb, group); 831 struct ext4_group_desc *gdp = ext4_get_group_desc(sb, group, NULL); 832 int ret; 833 834 if (flags & EXT4_GROUP_INFO_BBITMAP_CORRUPT) { 835 ret = ext4_test_and_set_bit(EXT4_GROUP_INFO_BBITMAP_CORRUPT_BIT, 836 &grp->bb_state); 837 if (!ret) 838 percpu_counter_sub(&sbi->s_freeclusters_counter, 839 grp->bb_free); 840 } 841 842 if (flags & EXT4_GROUP_INFO_IBITMAP_CORRUPT) { 843 ret = ext4_test_and_set_bit(EXT4_GROUP_INFO_IBITMAP_CORRUPT_BIT, 844 &grp->bb_state); 845 if (!ret && gdp) { 846 int count; 847 848 count = ext4_free_inodes_count(sb, gdp); 849 percpu_counter_sub(&sbi->s_freeinodes_counter, 850 count); 851 } 852 } 853 } 854 855 void ext4_update_dynamic_rev(struct super_block *sb) 856 { 857 struct ext4_super_block *es = EXT4_SB(sb)->s_es; 858 859 if (le32_to_cpu(es->s_rev_level) > EXT4_GOOD_OLD_REV) 860 return; 861 862 ext4_warning(sb, 863 "updating to rev %d because of new feature flag, " 864 "running e2fsck is recommended", 865 EXT4_DYNAMIC_REV); 866 867 es->s_first_ino = cpu_to_le32(EXT4_GOOD_OLD_FIRST_INO); 868 es->s_inode_size = cpu_to_le16(EXT4_GOOD_OLD_INODE_SIZE); 869 es->s_rev_level = cpu_to_le32(EXT4_DYNAMIC_REV); 870 /* leave es->s_feature_*compat flags alone */ 871 /* es->s_uuid will be set by e2fsck if empty */ 872 873 /* 874 * The rest of the superblock fields should be zero, and if not it 875 * means they are likely already in use, so leave them alone. We 876 * can leave it up to e2fsck to clean up any inconsistencies there. 877 */ 878 } 879 880 /* 881 * Open the external journal device 882 */ 883 static struct block_device *ext4_blkdev_get(dev_t dev, struct super_block *sb) 884 { 885 struct block_device *bdev; 886 char b[BDEVNAME_SIZE]; 887 888 bdev = blkdev_get_by_dev(dev, FMODE_READ|FMODE_WRITE|FMODE_EXCL, sb); 889 if (IS_ERR(bdev)) 890 goto fail; 891 return bdev; 892 893 fail: 894 ext4_msg(sb, KERN_ERR, "failed to open journal device %s: %ld", 895 __bdevname(dev, b), PTR_ERR(bdev)); 896 return NULL; 897 } 898 899 /* 900 * Release the journal device 901 */ 902 static void ext4_blkdev_put(struct block_device *bdev) 903 { 904 blkdev_put(bdev, FMODE_READ|FMODE_WRITE|FMODE_EXCL); 905 } 906 907 static void ext4_blkdev_remove(struct ext4_sb_info *sbi) 908 { 909 struct block_device *bdev; 910 bdev = sbi->journal_bdev; 911 if (bdev) { 912 ext4_blkdev_put(bdev); 913 sbi->journal_bdev = NULL; 914 } 915 } 916 917 static inline struct inode *orphan_list_entry(struct list_head *l) 918 { 919 return &list_entry(l, struct ext4_inode_info, i_orphan)->vfs_inode; 920 } 921 922 static void dump_orphan_list(struct super_block *sb, struct ext4_sb_info *sbi) 923 { 924 struct list_head *l; 925 926 ext4_msg(sb, KERN_ERR, "sb orphan head is %d", 927 le32_to_cpu(sbi->s_es->s_last_orphan)); 928 929 printk(KERN_ERR "sb_info orphan list:\n"); 930 list_for_each(l, &sbi->s_orphan) { 931 struct inode *inode = orphan_list_entry(l); 932 printk(KERN_ERR " " 933 "inode %s:%lu at %p: mode %o, nlink %d, next %d\n", 934 inode->i_sb->s_id, inode->i_ino, inode, 935 inode->i_mode, inode->i_nlink, 936 NEXT_ORPHAN(inode)); 937 } 938 } 939 940 #ifdef CONFIG_QUOTA 941 static int ext4_quota_off(struct super_block *sb, int type); 942 943 static inline void ext4_quota_off_umount(struct super_block *sb) 944 { 945 int type; 946 947 /* Use our quota_off function to clear inode flags etc. */ 948 for (type = 0; type < EXT4_MAXQUOTAS; type++) 949 ext4_quota_off(sb, type); 950 } 951 952 /* 953 * This is a helper function which is used in the mount/remount 954 * codepaths (which holds s_umount) to fetch the quota file name. 955 */ 956 static inline char *get_qf_name(struct super_block *sb, 957 struct ext4_sb_info *sbi, 958 int type) 959 { 960 return rcu_dereference_protected(sbi->s_qf_names[type], 961 lockdep_is_held(&sb->s_umount)); 962 } 963 #else 964 static inline void ext4_quota_off_umount(struct super_block *sb) 965 { 966 } 967 #endif 968 969 static void ext4_put_super(struct super_block *sb) 970 { 971 struct ext4_sb_info *sbi = EXT4_SB(sb); 972 struct ext4_super_block *es = sbi->s_es; 973 int aborted = 0; 974 int i, err; 975 976 ext4_unregister_li_request(sb); 977 ext4_quota_off_umount(sb); 978 979 destroy_workqueue(sbi->rsv_conversion_wq); 980 981 if (sbi->s_journal) { 982 aborted = is_journal_aborted(sbi->s_journal); 983 err = jbd2_journal_destroy(sbi->s_journal); 984 sbi->s_journal = NULL; 985 if ((err < 0) && !aborted) 986 ext4_abort(sb, "Couldn't clean up the journal"); 987 } 988 989 ext4_unregister_sysfs(sb); 990 ext4_es_unregister_shrinker(sbi); 991 del_timer_sync(&sbi->s_err_report); 992 ext4_release_system_zone(sb); 993 ext4_mb_release(sb); 994 ext4_ext_release(sb); 995 996 if (!sb_rdonly(sb) && !aborted) { 997 ext4_clear_feature_journal_needs_recovery(sb); 998 es->s_state = cpu_to_le16(sbi->s_mount_state); 999 } 1000 if (!sb_rdonly(sb)) 1001 ext4_commit_super(sb, 1); 1002 1003 for (i = 0; i < sbi->s_gdb_count; i++) 1004 brelse(sbi->s_group_desc[i]); 1005 kvfree(sbi->s_group_desc); 1006 kvfree(sbi->s_flex_groups); 1007 percpu_counter_destroy(&sbi->s_freeclusters_counter); 1008 percpu_counter_destroy(&sbi->s_freeinodes_counter); 1009 percpu_counter_destroy(&sbi->s_dirs_counter); 1010 percpu_counter_destroy(&sbi->s_dirtyclusters_counter); 1011 percpu_free_rwsem(&sbi->s_journal_flag_rwsem); 1012 #ifdef CONFIG_QUOTA 1013 for (i = 0; i < EXT4_MAXQUOTAS; i++) 1014 kfree(get_qf_name(sb, sbi, i)); 1015 #endif 1016 1017 /* Debugging code just in case the in-memory inode orphan list 1018 * isn't empty. The on-disk one can be non-empty if we've 1019 * detected an error and taken the fs readonly, but the 1020 * in-memory list had better be clean by this point. */ 1021 if (!list_empty(&sbi->s_orphan)) 1022 dump_orphan_list(sb, sbi); 1023 J_ASSERT(list_empty(&sbi->s_orphan)); 1024 1025 sync_blockdev(sb->s_bdev); 1026 invalidate_bdev(sb->s_bdev); 1027 if (sbi->journal_bdev && sbi->journal_bdev != sb->s_bdev) { 1028 /* 1029 * Invalidate the journal device's buffers. We don't want them 1030 * floating about in memory - the physical journal device may 1031 * hotswapped, and it breaks the `ro-after' testing code. 1032 */ 1033 sync_blockdev(sbi->journal_bdev); 1034 invalidate_bdev(sbi->journal_bdev); 1035 ext4_blkdev_remove(sbi); 1036 } 1037 1038 ext4_xattr_destroy_cache(sbi->s_ea_inode_cache); 1039 sbi->s_ea_inode_cache = NULL; 1040 1041 ext4_xattr_destroy_cache(sbi->s_ea_block_cache); 1042 sbi->s_ea_block_cache = NULL; 1043 1044 if (sbi->s_mmp_tsk) 1045 kthread_stop(sbi->s_mmp_tsk); 1046 brelse(sbi->s_sbh); 1047 sb->s_fs_info = NULL; 1048 /* 1049 * Now that we are completely done shutting down the 1050 * superblock, we need to actually destroy the kobject. 1051 */ 1052 kobject_put(&sbi->s_kobj); 1053 wait_for_completion(&sbi->s_kobj_unregister); 1054 if (sbi->s_chksum_driver) 1055 crypto_free_shash(sbi->s_chksum_driver); 1056 kfree(sbi->s_blockgroup_lock); 1057 fs_put_dax(sbi->s_daxdev); 1058 #ifdef CONFIG_UNICODE 1059 utf8_unload(sbi->s_encoding); 1060 #endif 1061 kfree(sbi); 1062 } 1063 1064 static struct kmem_cache *ext4_inode_cachep; 1065 1066 /* 1067 * Called inside transaction, so use GFP_NOFS 1068 */ 1069 static struct inode *ext4_alloc_inode(struct super_block *sb) 1070 { 1071 struct ext4_inode_info *ei; 1072 1073 ei = kmem_cache_alloc(ext4_inode_cachep, GFP_NOFS); 1074 if (!ei) 1075 return NULL; 1076 1077 inode_set_iversion(&ei->vfs_inode, 1); 1078 spin_lock_init(&ei->i_raw_lock); 1079 INIT_LIST_HEAD(&ei->i_prealloc_list); 1080 spin_lock_init(&ei->i_prealloc_lock); 1081 ext4_es_init_tree(&ei->i_es_tree); 1082 rwlock_init(&ei->i_es_lock); 1083 INIT_LIST_HEAD(&ei->i_es_list); 1084 ei->i_es_all_nr = 0; 1085 ei->i_es_shk_nr = 0; 1086 ei->i_es_shrink_lblk = 0; 1087 ei->i_reserved_data_blocks = 0; 1088 ei->i_da_metadata_calc_len = 0; 1089 ei->i_da_metadata_calc_last_lblock = 0; 1090 spin_lock_init(&(ei->i_block_reservation_lock)); 1091 ext4_init_pending_tree(&ei->i_pending_tree); 1092 #ifdef CONFIG_QUOTA 1093 ei->i_reserved_quota = 0; 1094 memset(&ei->i_dquot, 0, sizeof(ei->i_dquot)); 1095 #endif 1096 ei->jinode = NULL; 1097 INIT_LIST_HEAD(&ei->i_rsv_conversion_list); 1098 spin_lock_init(&ei->i_completed_io_lock); 1099 ei->i_sync_tid = 0; 1100 ei->i_datasync_tid = 0; 1101 atomic_set(&ei->i_unwritten, 0); 1102 INIT_WORK(&ei->i_rsv_conversion_work, ext4_end_io_rsv_work); 1103 return &ei->vfs_inode; 1104 } 1105 1106 static int ext4_drop_inode(struct inode *inode) 1107 { 1108 int drop = generic_drop_inode(inode); 1109 1110 if (!drop) 1111 drop = fscrypt_drop_inode(inode); 1112 1113 trace_ext4_drop_inode(inode, drop); 1114 return drop; 1115 } 1116 1117 static void ext4_free_in_core_inode(struct inode *inode) 1118 { 1119 fscrypt_free_inode(inode); 1120 kmem_cache_free(ext4_inode_cachep, EXT4_I(inode)); 1121 } 1122 1123 static void ext4_destroy_inode(struct inode *inode) 1124 { 1125 if (!list_empty(&(EXT4_I(inode)->i_orphan))) { 1126 ext4_msg(inode->i_sb, KERN_ERR, 1127 "Inode %lu (%p): orphan list check failed!", 1128 inode->i_ino, EXT4_I(inode)); 1129 print_hex_dump(KERN_INFO, "", DUMP_PREFIX_ADDRESS, 16, 4, 1130 EXT4_I(inode), sizeof(struct ext4_inode_info), 1131 true); 1132 dump_stack(); 1133 } 1134 } 1135 1136 static void init_once(void *foo) 1137 { 1138 struct ext4_inode_info *ei = (struct ext4_inode_info *) foo; 1139 1140 INIT_LIST_HEAD(&ei->i_orphan); 1141 init_rwsem(&ei->xattr_sem); 1142 init_rwsem(&ei->i_data_sem); 1143 init_rwsem(&ei->i_mmap_sem); 1144 inode_init_once(&ei->vfs_inode); 1145 } 1146 1147 static int __init init_inodecache(void) 1148 { 1149 ext4_inode_cachep = kmem_cache_create_usercopy("ext4_inode_cache", 1150 sizeof(struct ext4_inode_info), 0, 1151 (SLAB_RECLAIM_ACCOUNT|SLAB_MEM_SPREAD| 1152 SLAB_ACCOUNT), 1153 offsetof(struct ext4_inode_info, i_data), 1154 sizeof_field(struct ext4_inode_info, i_data), 1155 init_once); 1156 if (ext4_inode_cachep == NULL) 1157 return -ENOMEM; 1158 return 0; 1159 } 1160 1161 static void destroy_inodecache(void) 1162 { 1163 /* 1164 * Make sure all delayed rcu free inodes are flushed before we 1165 * destroy cache. 1166 */ 1167 rcu_barrier(); 1168 kmem_cache_destroy(ext4_inode_cachep); 1169 } 1170 1171 void ext4_clear_inode(struct inode *inode) 1172 { 1173 invalidate_inode_buffers(inode); 1174 clear_inode(inode); 1175 ext4_discard_preallocations(inode); 1176 ext4_es_remove_extent(inode, 0, EXT_MAX_BLOCKS); 1177 dquot_drop(inode); 1178 if (EXT4_I(inode)->jinode) { 1179 jbd2_journal_release_jbd_inode(EXT4_JOURNAL(inode), 1180 EXT4_I(inode)->jinode); 1181 jbd2_free_inode(EXT4_I(inode)->jinode); 1182 EXT4_I(inode)->jinode = NULL; 1183 } 1184 fscrypt_put_encryption_info(inode); 1185 fsverity_cleanup_inode(inode); 1186 } 1187 1188 static struct inode *ext4_nfs_get_inode(struct super_block *sb, 1189 u64 ino, u32 generation) 1190 { 1191 struct inode *inode; 1192 1193 /* 1194 * Currently we don't know the generation for parent directory, so 1195 * a generation of 0 means "accept any" 1196 */ 1197 inode = ext4_iget(sb, ino, EXT4_IGET_HANDLE); 1198 if (IS_ERR(inode)) 1199 return ERR_CAST(inode); 1200 if (generation && inode->i_generation != generation) { 1201 iput(inode); 1202 return ERR_PTR(-ESTALE); 1203 } 1204 1205 return inode; 1206 } 1207 1208 static struct dentry *ext4_fh_to_dentry(struct super_block *sb, struct fid *fid, 1209 int fh_len, int fh_type) 1210 { 1211 return generic_fh_to_dentry(sb, fid, fh_len, fh_type, 1212 ext4_nfs_get_inode); 1213 } 1214 1215 static struct dentry *ext4_fh_to_parent(struct super_block *sb, struct fid *fid, 1216 int fh_len, int fh_type) 1217 { 1218 return generic_fh_to_parent(sb, fid, fh_len, fh_type, 1219 ext4_nfs_get_inode); 1220 } 1221 1222 static int ext4_nfs_commit_metadata(struct inode *inode) 1223 { 1224 struct writeback_control wbc = { 1225 .sync_mode = WB_SYNC_ALL 1226 }; 1227 1228 trace_ext4_nfs_commit_metadata(inode); 1229 return ext4_write_inode(inode, &wbc); 1230 } 1231 1232 /* 1233 * Try to release metadata pages (indirect blocks, directories) which are 1234 * mapped via the block device. Since these pages could have journal heads 1235 * which would prevent try_to_free_buffers() from freeing them, we must use 1236 * jbd2 layer's try_to_free_buffers() function to release them. 1237 */ 1238 static int bdev_try_to_free_page(struct super_block *sb, struct page *page, 1239 gfp_t wait) 1240 { 1241 journal_t *journal = EXT4_SB(sb)->s_journal; 1242 1243 WARN_ON(PageChecked(page)); 1244 if (!page_has_buffers(page)) 1245 return 0; 1246 if (journal) 1247 return jbd2_journal_try_to_free_buffers(journal, page, 1248 wait & ~__GFP_DIRECT_RECLAIM); 1249 return try_to_free_buffers(page); 1250 } 1251 1252 #ifdef CONFIG_FS_ENCRYPTION 1253 static int ext4_get_context(struct inode *inode, void *ctx, size_t len) 1254 { 1255 return ext4_xattr_get(inode, EXT4_XATTR_INDEX_ENCRYPTION, 1256 EXT4_XATTR_NAME_ENCRYPTION_CONTEXT, ctx, len); 1257 } 1258 1259 static int ext4_set_context(struct inode *inode, const void *ctx, size_t len, 1260 void *fs_data) 1261 { 1262 handle_t *handle = fs_data; 1263 int res, res2, credits, retries = 0; 1264 1265 /* 1266 * Encrypting the root directory is not allowed because e2fsck expects 1267 * lost+found to exist and be unencrypted, and encrypting the root 1268 * directory would imply encrypting the lost+found directory as well as 1269 * the filename "lost+found" itself. 1270 */ 1271 if (inode->i_ino == EXT4_ROOT_INO) 1272 return -EPERM; 1273 1274 if (WARN_ON_ONCE(IS_DAX(inode) && i_size_read(inode))) 1275 return -EINVAL; 1276 1277 res = ext4_convert_inline_data(inode); 1278 if (res) 1279 return res; 1280 1281 /* 1282 * If a journal handle was specified, then the encryption context is 1283 * being set on a new inode via inheritance and is part of a larger 1284 * transaction to create the inode. Otherwise the encryption context is 1285 * being set on an existing inode in its own transaction. Only in the 1286 * latter case should the "retry on ENOSPC" logic be used. 1287 */ 1288 1289 if (handle) { 1290 res = ext4_xattr_set_handle(handle, inode, 1291 EXT4_XATTR_INDEX_ENCRYPTION, 1292 EXT4_XATTR_NAME_ENCRYPTION_CONTEXT, 1293 ctx, len, 0); 1294 if (!res) { 1295 ext4_set_inode_flag(inode, EXT4_INODE_ENCRYPT); 1296 ext4_clear_inode_state(inode, 1297 EXT4_STATE_MAY_INLINE_DATA); 1298 /* 1299 * Update inode->i_flags - S_ENCRYPTED will be enabled, 1300 * S_DAX may be disabled 1301 */ 1302 ext4_set_inode_flags(inode); 1303 } 1304 return res; 1305 } 1306 1307 res = dquot_initialize(inode); 1308 if (res) 1309 return res; 1310 retry: 1311 res = ext4_xattr_set_credits(inode, len, false /* is_create */, 1312 &credits); 1313 if (res) 1314 return res; 1315 1316 handle = ext4_journal_start(inode, EXT4_HT_MISC, credits); 1317 if (IS_ERR(handle)) 1318 return PTR_ERR(handle); 1319 1320 res = ext4_xattr_set_handle(handle, inode, EXT4_XATTR_INDEX_ENCRYPTION, 1321 EXT4_XATTR_NAME_ENCRYPTION_CONTEXT, 1322 ctx, len, 0); 1323 if (!res) { 1324 ext4_set_inode_flag(inode, EXT4_INODE_ENCRYPT); 1325 /* 1326 * Update inode->i_flags - S_ENCRYPTED will be enabled, 1327 * S_DAX may be disabled 1328 */ 1329 ext4_set_inode_flags(inode); 1330 res = ext4_mark_inode_dirty(handle, inode); 1331 if (res) 1332 EXT4_ERROR_INODE(inode, "Failed to mark inode dirty"); 1333 } 1334 res2 = ext4_journal_stop(handle); 1335 1336 if (res == -ENOSPC && ext4_should_retry_alloc(inode->i_sb, &retries)) 1337 goto retry; 1338 if (!res) 1339 res = res2; 1340 return res; 1341 } 1342 1343 static bool ext4_dummy_context(struct inode *inode) 1344 { 1345 return DUMMY_ENCRYPTION_ENABLED(EXT4_SB(inode->i_sb)); 1346 } 1347 1348 static bool ext4_has_stable_inodes(struct super_block *sb) 1349 { 1350 return ext4_has_feature_stable_inodes(sb); 1351 } 1352 1353 static void ext4_get_ino_and_lblk_bits(struct super_block *sb, 1354 int *ino_bits_ret, int *lblk_bits_ret) 1355 { 1356 *ino_bits_ret = 8 * sizeof(EXT4_SB(sb)->s_es->s_inodes_count); 1357 *lblk_bits_ret = 8 * sizeof(ext4_lblk_t); 1358 } 1359 1360 static const struct fscrypt_operations ext4_cryptops = { 1361 .key_prefix = "ext4:", 1362 .get_context = ext4_get_context, 1363 .set_context = ext4_set_context, 1364 .dummy_context = ext4_dummy_context, 1365 .empty_dir = ext4_empty_dir, 1366 .max_namelen = EXT4_NAME_LEN, 1367 .has_stable_inodes = ext4_has_stable_inodes, 1368 .get_ino_and_lblk_bits = ext4_get_ino_and_lblk_bits, 1369 }; 1370 #endif 1371 1372 #ifdef CONFIG_QUOTA 1373 static const char * const quotatypes[] = INITQFNAMES; 1374 #define QTYPE2NAME(t) (quotatypes[t]) 1375 1376 static int ext4_write_dquot(struct dquot *dquot); 1377 static int ext4_acquire_dquot(struct dquot *dquot); 1378 static int ext4_release_dquot(struct dquot *dquot); 1379 static int ext4_mark_dquot_dirty(struct dquot *dquot); 1380 static int ext4_write_info(struct super_block *sb, int type); 1381 static int ext4_quota_on(struct super_block *sb, int type, int format_id, 1382 const struct path *path); 1383 static int ext4_quota_on_mount(struct super_block *sb, int type); 1384 static ssize_t ext4_quota_read(struct super_block *sb, int type, char *data, 1385 size_t len, loff_t off); 1386 static ssize_t ext4_quota_write(struct super_block *sb, int type, 1387 const char *data, size_t len, loff_t off); 1388 static int ext4_quota_enable(struct super_block *sb, int type, int format_id, 1389 unsigned int flags); 1390 static int ext4_enable_quotas(struct super_block *sb); 1391 1392 static struct dquot **ext4_get_dquots(struct inode *inode) 1393 { 1394 return EXT4_I(inode)->i_dquot; 1395 } 1396 1397 static const struct dquot_operations ext4_quota_operations = { 1398 .get_reserved_space = ext4_get_reserved_space, 1399 .write_dquot = ext4_write_dquot, 1400 .acquire_dquot = ext4_acquire_dquot, 1401 .release_dquot = ext4_release_dquot, 1402 .mark_dirty = ext4_mark_dquot_dirty, 1403 .write_info = ext4_write_info, 1404 .alloc_dquot = dquot_alloc, 1405 .destroy_dquot = dquot_destroy, 1406 .get_projid = ext4_get_projid, 1407 .get_inode_usage = ext4_get_inode_usage, 1408 .get_next_id = dquot_get_next_id, 1409 }; 1410 1411 static const struct quotactl_ops ext4_qctl_operations = { 1412 .quota_on = ext4_quota_on, 1413 .quota_off = ext4_quota_off, 1414 .quota_sync = dquot_quota_sync, 1415 .get_state = dquot_get_state, 1416 .set_info = dquot_set_dqinfo, 1417 .get_dqblk = dquot_get_dqblk, 1418 .set_dqblk = dquot_set_dqblk, 1419 .get_nextdqblk = dquot_get_next_dqblk, 1420 }; 1421 #endif 1422 1423 static const struct super_operations ext4_sops = { 1424 .alloc_inode = ext4_alloc_inode, 1425 .free_inode = ext4_free_in_core_inode, 1426 .destroy_inode = ext4_destroy_inode, 1427 .write_inode = ext4_write_inode, 1428 .dirty_inode = ext4_dirty_inode, 1429 .drop_inode = ext4_drop_inode, 1430 .evict_inode = ext4_evict_inode, 1431 .put_super = ext4_put_super, 1432 .sync_fs = ext4_sync_fs, 1433 .freeze_fs = ext4_freeze, 1434 .unfreeze_fs = ext4_unfreeze, 1435 .statfs = ext4_statfs, 1436 .remount_fs = ext4_remount, 1437 .show_options = ext4_show_options, 1438 #ifdef CONFIG_QUOTA 1439 .quota_read = ext4_quota_read, 1440 .quota_write = ext4_quota_write, 1441 .get_dquots = ext4_get_dquots, 1442 #endif 1443 .bdev_try_to_free_page = bdev_try_to_free_page, 1444 }; 1445 1446 static const struct export_operations ext4_export_ops = { 1447 .fh_to_dentry = ext4_fh_to_dentry, 1448 .fh_to_parent = ext4_fh_to_parent, 1449 .get_parent = ext4_get_parent, 1450 .commit_metadata = ext4_nfs_commit_metadata, 1451 }; 1452 1453 enum { 1454 Opt_bsd_df, Opt_minix_df, Opt_grpid, Opt_nogrpid, 1455 Opt_resgid, Opt_resuid, Opt_sb, Opt_err_cont, Opt_err_panic, Opt_err_ro, 1456 Opt_nouid32, Opt_debug, Opt_removed, 1457 Opt_user_xattr, Opt_nouser_xattr, Opt_acl, Opt_noacl, 1458 Opt_auto_da_alloc, Opt_noauto_da_alloc, Opt_noload, 1459 Opt_commit, Opt_min_batch_time, Opt_max_batch_time, Opt_journal_dev, 1460 Opt_journal_path, Opt_journal_checksum, Opt_journal_async_commit, 1461 Opt_abort, Opt_data_journal, Opt_data_ordered, Opt_data_writeback, 1462 Opt_data_err_abort, Opt_data_err_ignore, Opt_test_dummy_encryption, 1463 Opt_usrjquota, Opt_grpjquota, Opt_offusrjquota, Opt_offgrpjquota, 1464 Opt_jqfmt_vfsold, Opt_jqfmt_vfsv0, Opt_jqfmt_vfsv1, Opt_quota, 1465 Opt_noquota, Opt_barrier, Opt_nobarrier, Opt_err, 1466 Opt_usrquota, Opt_grpquota, Opt_prjquota, Opt_i_version, Opt_dax, 1467 Opt_stripe, Opt_delalloc, Opt_nodelalloc, Opt_warn_on_error, 1468 Opt_nowarn_on_error, Opt_mblk_io_submit, 1469 Opt_lazytime, Opt_nolazytime, Opt_debug_want_extra_isize, 1470 Opt_nomblk_io_submit, Opt_block_validity, Opt_noblock_validity, 1471 Opt_inode_readahead_blks, Opt_journal_ioprio, 1472 Opt_dioread_nolock, Opt_dioread_lock, 1473 Opt_discard, Opt_nodiscard, Opt_init_itable, Opt_noinit_itable, 1474 Opt_max_dir_size_kb, Opt_nojournal_checksum, Opt_nombcache, 1475 }; 1476 1477 static const match_table_t tokens = { 1478 {Opt_bsd_df, "bsddf"}, 1479 {Opt_minix_df, "minixdf"}, 1480 {Opt_grpid, "grpid"}, 1481 {Opt_grpid, "bsdgroups"}, 1482 {Opt_nogrpid, "nogrpid"}, 1483 {Opt_nogrpid, "sysvgroups"}, 1484 {Opt_resgid, "resgid=%u"}, 1485 {Opt_resuid, "resuid=%u"}, 1486 {Opt_sb, "sb=%u"}, 1487 {Opt_err_cont, "errors=continue"}, 1488 {Opt_err_panic, "errors=panic"}, 1489 {Opt_err_ro, "errors=remount-ro"}, 1490 {Opt_nouid32, "nouid32"}, 1491 {Opt_debug, "debug"}, 1492 {Opt_removed, "oldalloc"}, 1493 {Opt_removed, "orlov"}, 1494 {Opt_user_xattr, "user_xattr"}, 1495 {Opt_nouser_xattr, "nouser_xattr"}, 1496 {Opt_acl, "acl"}, 1497 {Opt_noacl, "noacl"}, 1498 {Opt_noload, "norecovery"}, 1499 {Opt_noload, "noload"}, 1500 {Opt_removed, "nobh"}, 1501 {Opt_removed, "bh"}, 1502 {Opt_commit, "commit=%u"}, 1503 {Opt_min_batch_time, "min_batch_time=%u"}, 1504 {Opt_max_batch_time, "max_batch_time=%u"}, 1505 {Opt_journal_dev, "journal_dev=%u"}, 1506 {Opt_journal_path, "journal_path=%s"}, 1507 {Opt_journal_checksum, "journal_checksum"}, 1508 {Opt_nojournal_checksum, "nojournal_checksum"}, 1509 {Opt_journal_async_commit, "journal_async_commit"}, 1510 {Opt_abort, "abort"}, 1511 {Opt_data_journal, "data=journal"}, 1512 {Opt_data_ordered, "data=ordered"}, 1513 {Opt_data_writeback, "data=writeback"}, 1514 {Opt_data_err_abort, "data_err=abort"}, 1515 {Opt_data_err_ignore, "data_err=ignore"}, 1516 {Opt_offusrjquota, "usrjquota="}, 1517 {Opt_usrjquota, "usrjquota=%s"}, 1518 {Opt_offgrpjquota, "grpjquota="}, 1519 {Opt_grpjquota, "grpjquota=%s"}, 1520 {Opt_jqfmt_vfsold, "jqfmt=vfsold"}, 1521 {Opt_jqfmt_vfsv0, "jqfmt=vfsv0"}, 1522 {Opt_jqfmt_vfsv1, "jqfmt=vfsv1"}, 1523 {Opt_grpquota, "grpquota"}, 1524 {Opt_noquota, "noquota"}, 1525 {Opt_quota, "quota"}, 1526 {Opt_usrquota, "usrquota"}, 1527 {Opt_prjquota, "prjquota"}, 1528 {Opt_barrier, "barrier=%u"}, 1529 {Opt_barrier, "barrier"}, 1530 {Opt_nobarrier, "nobarrier"}, 1531 {Opt_i_version, "i_version"}, 1532 {Opt_dax, "dax"}, 1533 {Opt_stripe, "stripe=%u"}, 1534 {Opt_delalloc, "delalloc"}, 1535 {Opt_warn_on_error, "warn_on_error"}, 1536 {Opt_nowarn_on_error, "nowarn_on_error"}, 1537 {Opt_lazytime, "lazytime"}, 1538 {Opt_nolazytime, "nolazytime"}, 1539 {Opt_debug_want_extra_isize, "debug_want_extra_isize=%u"}, 1540 {Opt_nodelalloc, "nodelalloc"}, 1541 {Opt_removed, "mblk_io_submit"}, 1542 {Opt_removed, "nomblk_io_submit"}, 1543 {Opt_block_validity, "block_validity"}, 1544 {Opt_noblock_validity, "noblock_validity"}, 1545 {Opt_inode_readahead_blks, "inode_readahead_blks=%u"}, 1546 {Opt_journal_ioprio, "journal_ioprio=%u"}, 1547 {Opt_auto_da_alloc, "auto_da_alloc=%u"}, 1548 {Opt_auto_da_alloc, "auto_da_alloc"}, 1549 {Opt_noauto_da_alloc, "noauto_da_alloc"}, 1550 {Opt_dioread_nolock, "dioread_nolock"}, 1551 {Opt_dioread_lock, "dioread_lock"}, 1552 {Opt_discard, "discard"}, 1553 {Opt_nodiscard, "nodiscard"}, 1554 {Opt_init_itable, "init_itable=%u"}, 1555 {Opt_init_itable, "init_itable"}, 1556 {Opt_noinit_itable, "noinit_itable"}, 1557 {Opt_max_dir_size_kb, "max_dir_size_kb=%u"}, 1558 {Opt_test_dummy_encryption, "test_dummy_encryption"}, 1559 {Opt_nombcache, "nombcache"}, 1560 {Opt_nombcache, "no_mbcache"}, /* for backward compatibility */ 1561 {Opt_removed, "check=none"}, /* mount option from ext2/3 */ 1562 {Opt_removed, "nocheck"}, /* mount option from ext2/3 */ 1563 {Opt_removed, "reservation"}, /* mount option from ext2/3 */ 1564 {Opt_removed, "noreservation"}, /* mount option from ext2/3 */ 1565 {Opt_removed, "journal=%u"}, /* mount option from ext2/3 */ 1566 {Opt_err, NULL}, 1567 }; 1568 1569 static ext4_fsblk_t get_sb_block(void **data) 1570 { 1571 ext4_fsblk_t sb_block; 1572 char *options = (char *) *data; 1573 1574 if (!options || strncmp(options, "sb=", 3) != 0) 1575 return 1; /* Default location */ 1576 1577 options += 3; 1578 /* TODO: use simple_strtoll with >32bit ext4 */ 1579 sb_block = simple_strtoul(options, &options, 0); 1580 if (*options && *options != ',') { 1581 printk(KERN_ERR "EXT4-fs: Invalid sb specification: %s\n", 1582 (char *) *data); 1583 return 1; 1584 } 1585 if (*options == ',') 1586 options++; 1587 *data = (void *) options; 1588 1589 return sb_block; 1590 } 1591 1592 #define DEFAULT_JOURNAL_IOPRIO (IOPRIO_PRIO_VALUE(IOPRIO_CLASS_BE, 3)) 1593 static const char deprecated_msg[] = 1594 "Mount option \"%s\" will be removed by %s\n" 1595 "Contact linux-ext4@vger.kernel.org if you think we should keep it.\n"; 1596 1597 #ifdef CONFIG_QUOTA 1598 static int set_qf_name(struct super_block *sb, int qtype, substring_t *args) 1599 { 1600 struct ext4_sb_info *sbi = EXT4_SB(sb); 1601 char *qname, *old_qname = get_qf_name(sb, sbi, qtype); 1602 int ret = -1; 1603 1604 if (sb_any_quota_loaded(sb) && !old_qname) { 1605 ext4_msg(sb, KERN_ERR, 1606 "Cannot change journaled " 1607 "quota options when quota turned on"); 1608 return -1; 1609 } 1610 if (ext4_has_feature_quota(sb)) { 1611 ext4_msg(sb, KERN_INFO, "Journaled quota options " 1612 "ignored when QUOTA feature is enabled"); 1613 return 1; 1614 } 1615 qname = match_strdup(args); 1616 if (!qname) { 1617 ext4_msg(sb, KERN_ERR, 1618 "Not enough memory for storing quotafile name"); 1619 return -1; 1620 } 1621 if (old_qname) { 1622 if (strcmp(old_qname, qname) == 0) 1623 ret = 1; 1624 else 1625 ext4_msg(sb, KERN_ERR, 1626 "%s quota file already specified", 1627 QTYPE2NAME(qtype)); 1628 goto errout; 1629 } 1630 if (strchr(qname, '/')) { 1631 ext4_msg(sb, KERN_ERR, 1632 "quotafile must be on filesystem root"); 1633 goto errout; 1634 } 1635 rcu_assign_pointer(sbi->s_qf_names[qtype], qname); 1636 set_opt(sb, QUOTA); 1637 return 1; 1638 errout: 1639 kfree(qname); 1640 return ret; 1641 } 1642 1643 static int clear_qf_name(struct super_block *sb, int qtype) 1644 { 1645 1646 struct ext4_sb_info *sbi = EXT4_SB(sb); 1647 char *old_qname = get_qf_name(sb, sbi, qtype); 1648 1649 if (sb_any_quota_loaded(sb) && old_qname) { 1650 ext4_msg(sb, KERN_ERR, "Cannot change journaled quota options" 1651 " when quota turned on"); 1652 return -1; 1653 } 1654 rcu_assign_pointer(sbi->s_qf_names[qtype], NULL); 1655 synchronize_rcu(); 1656 kfree(old_qname); 1657 return 1; 1658 } 1659 #endif 1660 1661 #define MOPT_SET 0x0001 1662 #define MOPT_CLEAR 0x0002 1663 #define MOPT_NOSUPPORT 0x0004 1664 #define MOPT_EXPLICIT 0x0008 1665 #define MOPT_CLEAR_ERR 0x0010 1666 #define MOPT_GTE0 0x0020 1667 #ifdef CONFIG_QUOTA 1668 #define MOPT_Q 0 1669 #define MOPT_QFMT 0x0040 1670 #else 1671 #define MOPT_Q MOPT_NOSUPPORT 1672 #define MOPT_QFMT MOPT_NOSUPPORT 1673 #endif 1674 #define MOPT_DATAJ 0x0080 1675 #define MOPT_NO_EXT2 0x0100 1676 #define MOPT_NO_EXT3 0x0200 1677 #define MOPT_EXT4_ONLY (MOPT_NO_EXT2 | MOPT_NO_EXT3) 1678 #define MOPT_STRING 0x0400 1679 1680 static const struct mount_opts { 1681 int token; 1682 int mount_opt; 1683 int flags; 1684 } ext4_mount_opts[] = { 1685 {Opt_minix_df, EXT4_MOUNT_MINIX_DF, MOPT_SET}, 1686 {Opt_bsd_df, EXT4_MOUNT_MINIX_DF, MOPT_CLEAR}, 1687 {Opt_grpid, EXT4_MOUNT_GRPID, MOPT_SET}, 1688 {Opt_nogrpid, EXT4_MOUNT_GRPID, MOPT_CLEAR}, 1689 {Opt_block_validity, EXT4_MOUNT_BLOCK_VALIDITY, MOPT_SET}, 1690 {Opt_noblock_validity, EXT4_MOUNT_BLOCK_VALIDITY, MOPT_CLEAR}, 1691 {Opt_dioread_nolock, EXT4_MOUNT_DIOREAD_NOLOCK, 1692 MOPT_EXT4_ONLY | MOPT_SET}, 1693 {Opt_dioread_lock, EXT4_MOUNT_DIOREAD_NOLOCK, 1694 MOPT_EXT4_ONLY | MOPT_CLEAR}, 1695 {Opt_discard, EXT4_MOUNT_DISCARD, MOPT_SET}, 1696 {Opt_nodiscard, EXT4_MOUNT_DISCARD, MOPT_CLEAR}, 1697 {Opt_delalloc, EXT4_MOUNT_DELALLOC, 1698 MOPT_EXT4_ONLY | MOPT_SET | MOPT_EXPLICIT}, 1699 {Opt_nodelalloc, EXT4_MOUNT_DELALLOC, 1700 MOPT_EXT4_ONLY | MOPT_CLEAR}, 1701 {Opt_warn_on_error, EXT4_MOUNT_WARN_ON_ERROR, MOPT_SET}, 1702 {Opt_nowarn_on_error, EXT4_MOUNT_WARN_ON_ERROR, MOPT_CLEAR}, 1703 {Opt_nojournal_checksum, EXT4_MOUNT_JOURNAL_CHECKSUM, 1704 MOPT_EXT4_ONLY | MOPT_CLEAR}, 1705 {Opt_journal_checksum, EXT4_MOUNT_JOURNAL_CHECKSUM, 1706 MOPT_EXT4_ONLY | MOPT_SET | MOPT_EXPLICIT}, 1707 {Opt_journal_async_commit, (EXT4_MOUNT_JOURNAL_ASYNC_COMMIT | 1708 EXT4_MOUNT_JOURNAL_CHECKSUM), 1709 MOPT_EXT4_ONLY | MOPT_SET | MOPT_EXPLICIT}, 1710 {Opt_noload, EXT4_MOUNT_NOLOAD, MOPT_NO_EXT2 | MOPT_SET}, 1711 {Opt_err_panic, EXT4_MOUNT_ERRORS_PANIC, MOPT_SET | MOPT_CLEAR_ERR}, 1712 {Opt_err_ro, EXT4_MOUNT_ERRORS_RO, MOPT_SET | MOPT_CLEAR_ERR}, 1713 {Opt_err_cont, EXT4_MOUNT_ERRORS_CONT, MOPT_SET | MOPT_CLEAR_ERR}, 1714 {Opt_data_err_abort, EXT4_MOUNT_DATA_ERR_ABORT, 1715 MOPT_NO_EXT2}, 1716 {Opt_data_err_ignore, EXT4_MOUNT_DATA_ERR_ABORT, 1717 MOPT_NO_EXT2}, 1718 {Opt_barrier, EXT4_MOUNT_BARRIER, MOPT_SET}, 1719 {Opt_nobarrier, EXT4_MOUNT_BARRIER, MOPT_CLEAR}, 1720 {Opt_noauto_da_alloc, EXT4_MOUNT_NO_AUTO_DA_ALLOC, MOPT_SET}, 1721 {Opt_auto_da_alloc, EXT4_MOUNT_NO_AUTO_DA_ALLOC, MOPT_CLEAR}, 1722 {Opt_noinit_itable, EXT4_MOUNT_INIT_INODE_TABLE, MOPT_CLEAR}, 1723 {Opt_commit, 0, MOPT_GTE0}, 1724 {Opt_max_batch_time, 0, MOPT_GTE0}, 1725 {Opt_min_batch_time, 0, MOPT_GTE0}, 1726 {Opt_inode_readahead_blks, 0, MOPT_GTE0}, 1727 {Opt_init_itable, 0, MOPT_GTE0}, 1728 {Opt_dax, EXT4_MOUNT_DAX, MOPT_SET}, 1729 {Opt_stripe, 0, MOPT_GTE0}, 1730 {Opt_resuid, 0, MOPT_GTE0}, 1731 {Opt_resgid, 0, MOPT_GTE0}, 1732 {Opt_journal_dev, 0, MOPT_NO_EXT2 | MOPT_GTE0}, 1733 {Opt_journal_path, 0, MOPT_NO_EXT2 | MOPT_STRING}, 1734 {Opt_journal_ioprio, 0, MOPT_NO_EXT2 | MOPT_GTE0}, 1735 {Opt_data_journal, EXT4_MOUNT_JOURNAL_DATA, MOPT_NO_EXT2 | MOPT_DATAJ}, 1736 {Opt_data_ordered, EXT4_MOUNT_ORDERED_DATA, MOPT_NO_EXT2 | MOPT_DATAJ}, 1737 {Opt_data_writeback, EXT4_MOUNT_WRITEBACK_DATA, 1738 MOPT_NO_EXT2 | MOPT_DATAJ}, 1739 {Opt_user_xattr, EXT4_MOUNT_XATTR_USER, MOPT_SET}, 1740 {Opt_nouser_xattr, EXT4_MOUNT_XATTR_USER, MOPT_CLEAR}, 1741 #ifdef CONFIG_EXT4_FS_POSIX_ACL 1742 {Opt_acl, EXT4_MOUNT_POSIX_ACL, MOPT_SET}, 1743 {Opt_noacl, EXT4_MOUNT_POSIX_ACL, MOPT_CLEAR}, 1744 #else 1745 {Opt_acl, 0, MOPT_NOSUPPORT}, 1746 {Opt_noacl, 0, MOPT_NOSUPPORT}, 1747 #endif 1748 {Opt_nouid32, EXT4_MOUNT_NO_UID32, MOPT_SET}, 1749 {Opt_debug, EXT4_MOUNT_DEBUG, MOPT_SET}, 1750 {Opt_debug_want_extra_isize, 0, MOPT_GTE0}, 1751 {Opt_quota, EXT4_MOUNT_QUOTA | EXT4_MOUNT_USRQUOTA, MOPT_SET | MOPT_Q}, 1752 {Opt_usrquota, EXT4_MOUNT_QUOTA | EXT4_MOUNT_USRQUOTA, 1753 MOPT_SET | MOPT_Q}, 1754 {Opt_grpquota, EXT4_MOUNT_QUOTA | EXT4_MOUNT_GRPQUOTA, 1755 MOPT_SET | MOPT_Q}, 1756 {Opt_prjquota, EXT4_MOUNT_QUOTA | EXT4_MOUNT_PRJQUOTA, 1757 MOPT_SET | MOPT_Q}, 1758 {Opt_noquota, (EXT4_MOUNT_QUOTA | EXT4_MOUNT_USRQUOTA | 1759 EXT4_MOUNT_GRPQUOTA | EXT4_MOUNT_PRJQUOTA), 1760 MOPT_CLEAR | MOPT_Q}, 1761 {Opt_usrjquota, 0, MOPT_Q}, 1762 {Opt_grpjquota, 0, MOPT_Q}, 1763 {Opt_offusrjquota, 0, MOPT_Q}, 1764 {Opt_offgrpjquota, 0, MOPT_Q}, 1765 {Opt_jqfmt_vfsold, QFMT_VFS_OLD, MOPT_QFMT}, 1766 {Opt_jqfmt_vfsv0, QFMT_VFS_V0, MOPT_QFMT}, 1767 {Opt_jqfmt_vfsv1, QFMT_VFS_V1, MOPT_QFMT}, 1768 {Opt_max_dir_size_kb, 0, MOPT_GTE0}, 1769 {Opt_test_dummy_encryption, 0, MOPT_GTE0}, 1770 {Opt_nombcache, EXT4_MOUNT_NO_MBCACHE, MOPT_SET}, 1771 {Opt_err, 0, 0} 1772 }; 1773 1774 #ifdef CONFIG_UNICODE 1775 static const struct ext4_sb_encodings { 1776 __u16 magic; 1777 char *name; 1778 char *version; 1779 } ext4_sb_encoding_map[] = { 1780 {EXT4_ENC_UTF8_12_1, "utf8", "12.1.0"}, 1781 }; 1782 1783 static int ext4_sb_read_encoding(const struct ext4_super_block *es, 1784 const struct ext4_sb_encodings **encoding, 1785 __u16 *flags) 1786 { 1787 __u16 magic = le16_to_cpu(es->s_encoding); 1788 int i; 1789 1790 for (i = 0; i < ARRAY_SIZE(ext4_sb_encoding_map); i++) 1791 if (magic == ext4_sb_encoding_map[i].magic) 1792 break; 1793 1794 if (i >= ARRAY_SIZE(ext4_sb_encoding_map)) 1795 return -EINVAL; 1796 1797 *encoding = &ext4_sb_encoding_map[i]; 1798 *flags = le16_to_cpu(es->s_encoding_flags); 1799 1800 return 0; 1801 } 1802 #endif 1803 1804 static int handle_mount_opt(struct super_block *sb, char *opt, int token, 1805 substring_t *args, unsigned long *journal_devnum, 1806 unsigned int *journal_ioprio, int is_remount) 1807 { 1808 struct ext4_sb_info *sbi = EXT4_SB(sb); 1809 const struct mount_opts *m; 1810 kuid_t uid; 1811 kgid_t gid; 1812 int arg = 0; 1813 1814 #ifdef CONFIG_QUOTA 1815 if (token == Opt_usrjquota) 1816 return set_qf_name(sb, USRQUOTA, &args[0]); 1817 else if (token == Opt_grpjquota) 1818 return set_qf_name(sb, GRPQUOTA, &args[0]); 1819 else if (token == Opt_offusrjquota) 1820 return clear_qf_name(sb, USRQUOTA); 1821 else if (token == Opt_offgrpjquota) 1822 return clear_qf_name(sb, GRPQUOTA); 1823 #endif 1824 switch (token) { 1825 case Opt_noacl: 1826 case Opt_nouser_xattr: 1827 ext4_msg(sb, KERN_WARNING, deprecated_msg, opt, "3.5"); 1828 break; 1829 case Opt_sb: 1830 return 1; /* handled by get_sb_block() */ 1831 case Opt_removed: 1832 ext4_msg(sb, KERN_WARNING, "Ignoring removed %s option", opt); 1833 return 1; 1834 case Opt_abort: 1835 sbi->s_mount_flags |= EXT4_MF_FS_ABORTED; 1836 return 1; 1837 case Opt_i_version: 1838 sb->s_flags |= SB_I_VERSION; 1839 return 1; 1840 case Opt_lazytime: 1841 sb->s_flags |= SB_LAZYTIME; 1842 return 1; 1843 case Opt_nolazytime: 1844 sb->s_flags &= ~SB_LAZYTIME; 1845 return 1; 1846 } 1847 1848 for (m = ext4_mount_opts; m->token != Opt_err; m++) 1849 if (token == m->token) 1850 break; 1851 1852 if (m->token == Opt_err) { 1853 ext4_msg(sb, KERN_ERR, "Unrecognized mount option \"%s\" " 1854 "or missing value", opt); 1855 return -1; 1856 } 1857 1858 if ((m->flags & MOPT_NO_EXT2) && IS_EXT2_SB(sb)) { 1859 ext4_msg(sb, KERN_ERR, 1860 "Mount option \"%s\" incompatible with ext2", opt); 1861 return -1; 1862 } 1863 if ((m->flags & MOPT_NO_EXT3) && IS_EXT3_SB(sb)) { 1864 ext4_msg(sb, KERN_ERR, 1865 "Mount option \"%s\" incompatible with ext3", opt); 1866 return -1; 1867 } 1868 1869 if (args->from && !(m->flags & MOPT_STRING) && match_int(args, &arg)) 1870 return -1; 1871 if (args->from && (m->flags & MOPT_GTE0) && (arg < 0)) 1872 return -1; 1873 if (m->flags & MOPT_EXPLICIT) { 1874 if (m->mount_opt & EXT4_MOUNT_DELALLOC) { 1875 set_opt2(sb, EXPLICIT_DELALLOC); 1876 } else if (m->mount_opt & EXT4_MOUNT_JOURNAL_CHECKSUM) { 1877 set_opt2(sb, EXPLICIT_JOURNAL_CHECKSUM); 1878 } else 1879 return -1; 1880 } 1881 if (m->flags & MOPT_CLEAR_ERR) 1882 clear_opt(sb, ERRORS_MASK); 1883 if (token == Opt_noquota && sb_any_quota_loaded(sb)) { 1884 ext4_msg(sb, KERN_ERR, "Cannot change quota " 1885 "options when quota turned on"); 1886 return -1; 1887 } 1888 1889 if (m->flags & MOPT_NOSUPPORT) { 1890 ext4_msg(sb, KERN_ERR, "%s option not supported", opt); 1891 } else if (token == Opt_commit) { 1892 if (arg == 0) 1893 arg = JBD2_DEFAULT_MAX_COMMIT_AGE; 1894 else if (arg > INT_MAX / HZ) { 1895 ext4_msg(sb, KERN_ERR, 1896 "Invalid commit interval %d, " 1897 "must be smaller than %d", 1898 arg, INT_MAX / HZ); 1899 return -1; 1900 } 1901 sbi->s_commit_interval = HZ * arg; 1902 } else if (token == Opt_debug_want_extra_isize) { 1903 if ((arg & 1) || 1904 (arg < 4) || 1905 (arg > (sbi->s_inode_size - EXT4_GOOD_OLD_INODE_SIZE))) { 1906 ext4_msg(sb, KERN_ERR, 1907 "Invalid want_extra_isize %d", arg); 1908 return -1; 1909 } 1910 sbi->s_want_extra_isize = arg; 1911 } else if (token == Opt_max_batch_time) { 1912 sbi->s_max_batch_time = arg; 1913 } else if (token == Opt_min_batch_time) { 1914 sbi->s_min_batch_time = arg; 1915 } else if (token == Opt_inode_readahead_blks) { 1916 if (arg && (arg > (1 << 30) || !is_power_of_2(arg))) { 1917 ext4_msg(sb, KERN_ERR, 1918 "EXT4-fs: inode_readahead_blks must be " 1919 "0 or a power of 2 smaller than 2^31"); 1920 return -1; 1921 } 1922 sbi->s_inode_readahead_blks = arg; 1923 } else if (token == Opt_init_itable) { 1924 set_opt(sb, INIT_INODE_TABLE); 1925 if (!args->from) 1926 arg = EXT4_DEF_LI_WAIT_MULT; 1927 sbi->s_li_wait_mult = arg; 1928 } else if (token == Opt_max_dir_size_kb) { 1929 sbi->s_max_dir_size_kb = arg; 1930 } else if (token == Opt_stripe) { 1931 sbi->s_stripe = arg; 1932 } else if (token == Opt_resuid) { 1933 uid = make_kuid(current_user_ns(), arg); 1934 if (!uid_valid(uid)) { 1935 ext4_msg(sb, KERN_ERR, "Invalid uid value %d", arg); 1936 return -1; 1937 } 1938 sbi->s_resuid = uid; 1939 } else if (token == Opt_resgid) { 1940 gid = make_kgid(current_user_ns(), arg); 1941 if (!gid_valid(gid)) { 1942 ext4_msg(sb, KERN_ERR, "Invalid gid value %d", arg); 1943 return -1; 1944 } 1945 sbi->s_resgid = gid; 1946 } else if (token == Opt_journal_dev) { 1947 if (is_remount) { 1948 ext4_msg(sb, KERN_ERR, 1949 "Cannot specify journal on remount"); 1950 return -1; 1951 } 1952 *journal_devnum = arg; 1953 } else if (token == Opt_journal_path) { 1954 char *journal_path; 1955 struct inode *journal_inode; 1956 struct path path; 1957 int error; 1958 1959 if (is_remount) { 1960 ext4_msg(sb, KERN_ERR, 1961 "Cannot specify journal on remount"); 1962 return -1; 1963 } 1964 journal_path = match_strdup(&args[0]); 1965 if (!journal_path) { 1966 ext4_msg(sb, KERN_ERR, "error: could not dup " 1967 "journal device string"); 1968 return -1; 1969 } 1970 1971 error = kern_path(journal_path, LOOKUP_FOLLOW, &path); 1972 if (error) { 1973 ext4_msg(sb, KERN_ERR, "error: could not find " 1974 "journal device path: error %d", error); 1975 kfree(journal_path); 1976 return -1; 1977 } 1978 1979 journal_inode = d_inode(path.dentry); 1980 if (!S_ISBLK(journal_inode->i_mode)) { 1981 ext4_msg(sb, KERN_ERR, "error: journal path %s " 1982 "is not a block device", journal_path); 1983 path_put(&path); 1984 kfree(journal_path); 1985 return -1; 1986 } 1987 1988 *journal_devnum = new_encode_dev(journal_inode->i_rdev); 1989 path_put(&path); 1990 kfree(journal_path); 1991 } else if (token == Opt_journal_ioprio) { 1992 if (arg > 7) { 1993 ext4_msg(sb, KERN_ERR, "Invalid journal IO priority" 1994 " (must be 0-7)"); 1995 return -1; 1996 } 1997 *journal_ioprio = 1998 IOPRIO_PRIO_VALUE(IOPRIO_CLASS_BE, arg); 1999 } else if (token == Opt_test_dummy_encryption) { 2000 #ifdef CONFIG_FS_ENCRYPTION 2001 sbi->s_mount_flags |= EXT4_MF_TEST_DUMMY_ENCRYPTION; 2002 ext4_msg(sb, KERN_WARNING, 2003 "Test dummy encryption mode enabled"); 2004 #else 2005 ext4_msg(sb, KERN_WARNING, 2006 "Test dummy encryption mount option ignored"); 2007 #endif 2008 } else if (m->flags & MOPT_DATAJ) { 2009 if (is_remount) { 2010 if (!sbi->s_journal) 2011 ext4_msg(sb, KERN_WARNING, "Remounting file system with no journal so ignoring journalled data option"); 2012 else if (test_opt(sb, DATA_FLAGS) != m->mount_opt) { 2013 ext4_msg(sb, KERN_ERR, 2014 "Cannot change data mode on remount"); 2015 return -1; 2016 } 2017 } else { 2018 clear_opt(sb, DATA_FLAGS); 2019 sbi->s_mount_opt |= m->mount_opt; 2020 } 2021 #ifdef CONFIG_QUOTA 2022 } else if (m->flags & MOPT_QFMT) { 2023 if (sb_any_quota_loaded(sb) && 2024 sbi->s_jquota_fmt != m->mount_opt) { 2025 ext4_msg(sb, KERN_ERR, "Cannot change journaled " 2026 "quota options when quota turned on"); 2027 return -1; 2028 } 2029 if (ext4_has_feature_quota(sb)) { 2030 ext4_msg(sb, KERN_INFO, 2031 "Quota format mount options ignored " 2032 "when QUOTA feature is enabled"); 2033 return 1; 2034 } 2035 sbi->s_jquota_fmt = m->mount_opt; 2036 #endif 2037 } else if (token == Opt_dax) { 2038 #ifdef CONFIG_FS_DAX 2039 ext4_msg(sb, KERN_WARNING, 2040 "DAX enabled. Warning: EXPERIMENTAL, use at your own risk"); 2041 sbi->s_mount_opt |= m->mount_opt; 2042 #else 2043 ext4_msg(sb, KERN_INFO, "dax option not supported"); 2044 return -1; 2045 #endif 2046 } else if (token == Opt_data_err_abort) { 2047 sbi->s_mount_opt |= m->mount_opt; 2048 } else if (token == Opt_data_err_ignore) { 2049 sbi->s_mount_opt &= ~m->mount_opt; 2050 } else { 2051 if (!args->from) 2052 arg = 1; 2053 if (m->flags & MOPT_CLEAR) 2054 arg = !arg; 2055 else if (unlikely(!(m->flags & MOPT_SET))) { 2056 ext4_msg(sb, KERN_WARNING, 2057 "buggy handling of option %s", opt); 2058 WARN_ON(1); 2059 return -1; 2060 } 2061 if (arg != 0) 2062 sbi->s_mount_opt |= m->mount_opt; 2063 else 2064 sbi->s_mount_opt &= ~m->mount_opt; 2065 } 2066 return 1; 2067 } 2068 2069 static int parse_options(char *options, struct super_block *sb, 2070 unsigned long *journal_devnum, 2071 unsigned int *journal_ioprio, 2072 int is_remount) 2073 { 2074 struct ext4_sb_info __maybe_unused *sbi = EXT4_SB(sb); 2075 char *p, __maybe_unused *usr_qf_name, __maybe_unused *grp_qf_name; 2076 substring_t args[MAX_OPT_ARGS]; 2077 int token; 2078 2079 if (!options) 2080 return 1; 2081 2082 while ((p = strsep(&options, ",")) != NULL) { 2083 if (!*p) 2084 continue; 2085 /* 2086 * Initialize args struct so we know whether arg was 2087 * found; some options take optional arguments. 2088 */ 2089 args[0].to = args[0].from = NULL; 2090 token = match_token(p, tokens, args); 2091 if (handle_mount_opt(sb, p, token, args, journal_devnum, 2092 journal_ioprio, is_remount) < 0) 2093 return 0; 2094 } 2095 #ifdef CONFIG_QUOTA 2096 /* 2097 * We do the test below only for project quotas. 'usrquota' and 2098 * 'grpquota' mount options are allowed even without quota feature 2099 * to support legacy quotas in quota files. 2100 */ 2101 if (test_opt(sb, PRJQUOTA) && !ext4_has_feature_project(sb)) { 2102 ext4_msg(sb, KERN_ERR, "Project quota feature not enabled. " 2103 "Cannot enable project quota enforcement."); 2104 return 0; 2105 } 2106 usr_qf_name = get_qf_name(sb, sbi, USRQUOTA); 2107 grp_qf_name = get_qf_name(sb, sbi, GRPQUOTA); 2108 if (usr_qf_name || grp_qf_name) { 2109 if (test_opt(sb, USRQUOTA) && usr_qf_name) 2110 clear_opt(sb, USRQUOTA); 2111 2112 if (test_opt(sb, GRPQUOTA) && grp_qf_name) 2113 clear_opt(sb, GRPQUOTA); 2114 2115 if (test_opt(sb, GRPQUOTA) || test_opt(sb, USRQUOTA)) { 2116 ext4_msg(sb, KERN_ERR, "old and new quota " 2117 "format mixing"); 2118 return 0; 2119 } 2120 2121 if (!sbi->s_jquota_fmt) { 2122 ext4_msg(sb, KERN_ERR, "journaled quota format " 2123 "not specified"); 2124 return 0; 2125 } 2126 } 2127 #endif 2128 return 1; 2129 } 2130 2131 static inline void ext4_show_quota_options(struct seq_file *seq, 2132 struct super_block *sb) 2133 { 2134 #if defined(CONFIG_QUOTA) 2135 struct ext4_sb_info *sbi = EXT4_SB(sb); 2136 char *usr_qf_name, *grp_qf_name; 2137 2138 if (sbi->s_jquota_fmt) { 2139 char *fmtname = ""; 2140 2141 switch (sbi->s_jquota_fmt) { 2142 case QFMT_VFS_OLD: 2143 fmtname = "vfsold"; 2144 break; 2145 case QFMT_VFS_V0: 2146 fmtname = "vfsv0"; 2147 break; 2148 case QFMT_VFS_V1: 2149 fmtname = "vfsv1"; 2150 break; 2151 } 2152 seq_printf(seq, ",jqfmt=%s", fmtname); 2153 } 2154 2155 rcu_read_lock(); 2156 usr_qf_name = rcu_dereference(sbi->s_qf_names[USRQUOTA]); 2157 grp_qf_name = rcu_dereference(sbi->s_qf_names[GRPQUOTA]); 2158 if (usr_qf_name) 2159 seq_show_option(seq, "usrjquota", usr_qf_name); 2160 if (grp_qf_name) 2161 seq_show_option(seq, "grpjquota", grp_qf_name); 2162 rcu_read_unlock(); 2163 #endif 2164 } 2165 2166 static const char *token2str(int token) 2167 { 2168 const struct match_token *t; 2169 2170 for (t = tokens; t->token != Opt_err; t++) 2171 if (t->token == token && !strchr(t->pattern, '=')) 2172 break; 2173 return t->pattern; 2174 } 2175 2176 /* 2177 * Show an option if 2178 * - it's set to a non-default value OR 2179 * - if the per-sb default is different from the global default 2180 */ 2181 static int _ext4_show_options(struct seq_file *seq, struct super_block *sb, 2182 int nodefs) 2183 { 2184 struct ext4_sb_info *sbi = EXT4_SB(sb); 2185 struct ext4_super_block *es = sbi->s_es; 2186 int def_errors, def_mount_opt = sbi->s_def_mount_opt; 2187 const struct mount_opts *m; 2188 char sep = nodefs ? '\n' : ','; 2189 2190 #define SEQ_OPTS_PUTS(str) seq_printf(seq, "%c" str, sep) 2191 #define SEQ_OPTS_PRINT(str, arg) seq_printf(seq, "%c" str, sep, arg) 2192 2193 if (sbi->s_sb_block != 1) 2194 SEQ_OPTS_PRINT("sb=%llu", sbi->s_sb_block); 2195 2196 for (m = ext4_mount_opts; m->token != Opt_err; m++) { 2197 int want_set = m->flags & MOPT_SET; 2198 if (((m->flags & (MOPT_SET|MOPT_CLEAR)) == 0) || 2199 (m->flags & MOPT_CLEAR_ERR)) 2200 continue; 2201 if (!nodefs && !(m->mount_opt & (sbi->s_mount_opt ^ def_mount_opt))) 2202 continue; /* skip if same as the default */ 2203 if ((want_set && 2204 (sbi->s_mount_opt & m->mount_opt) != m->mount_opt) || 2205 (!want_set && (sbi->s_mount_opt & m->mount_opt))) 2206 continue; /* select Opt_noFoo vs Opt_Foo */ 2207 SEQ_OPTS_PRINT("%s", token2str(m->token)); 2208 } 2209 2210 if (nodefs || !uid_eq(sbi->s_resuid, make_kuid(&init_user_ns, EXT4_DEF_RESUID)) || 2211 le16_to_cpu(es->s_def_resuid) != EXT4_DEF_RESUID) 2212 SEQ_OPTS_PRINT("resuid=%u", 2213 from_kuid_munged(&init_user_ns, sbi->s_resuid)); 2214 if (nodefs || !gid_eq(sbi->s_resgid, make_kgid(&init_user_ns, EXT4_DEF_RESGID)) || 2215 le16_to_cpu(es->s_def_resgid) != EXT4_DEF_RESGID) 2216 SEQ_OPTS_PRINT("resgid=%u", 2217 from_kgid_munged(&init_user_ns, sbi->s_resgid)); 2218 def_errors = nodefs ? -1 : le16_to_cpu(es->s_errors); 2219 if (test_opt(sb, ERRORS_RO) && def_errors != EXT4_ERRORS_RO) 2220 SEQ_OPTS_PUTS("errors=remount-ro"); 2221 if (test_opt(sb, ERRORS_CONT) && def_errors != EXT4_ERRORS_CONTINUE) 2222 SEQ_OPTS_PUTS("errors=continue"); 2223 if (test_opt(sb, ERRORS_PANIC) && def_errors != EXT4_ERRORS_PANIC) 2224 SEQ_OPTS_PUTS("errors=panic"); 2225 if (nodefs || sbi->s_commit_interval != JBD2_DEFAULT_MAX_COMMIT_AGE*HZ) 2226 SEQ_OPTS_PRINT("commit=%lu", sbi->s_commit_interval / HZ); 2227 if (nodefs || sbi->s_min_batch_time != EXT4_DEF_MIN_BATCH_TIME) 2228 SEQ_OPTS_PRINT("min_batch_time=%u", sbi->s_min_batch_time); 2229 if (nodefs || sbi->s_max_batch_time != EXT4_DEF_MAX_BATCH_TIME) 2230 SEQ_OPTS_PRINT("max_batch_time=%u", sbi->s_max_batch_time); 2231 if (sb->s_flags & SB_I_VERSION) 2232 SEQ_OPTS_PUTS("i_version"); 2233 if (nodefs || sbi->s_stripe) 2234 SEQ_OPTS_PRINT("stripe=%lu", sbi->s_stripe); 2235 if (nodefs || EXT4_MOUNT_DATA_FLAGS & 2236 (sbi->s_mount_opt ^ def_mount_opt)) { 2237 if (test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_JOURNAL_DATA) 2238 SEQ_OPTS_PUTS("data=journal"); 2239 else if (test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_ORDERED_DATA) 2240 SEQ_OPTS_PUTS("data=ordered"); 2241 else if (test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_WRITEBACK_DATA) 2242 SEQ_OPTS_PUTS("data=writeback"); 2243 } 2244 if (nodefs || 2245 sbi->s_inode_readahead_blks != EXT4_DEF_INODE_READAHEAD_BLKS) 2246 SEQ_OPTS_PRINT("inode_readahead_blks=%u", 2247 sbi->s_inode_readahead_blks); 2248 2249 if (test_opt(sb, INIT_INODE_TABLE) && (nodefs || 2250 (sbi->s_li_wait_mult != EXT4_DEF_LI_WAIT_MULT))) 2251 SEQ_OPTS_PRINT("init_itable=%u", sbi->s_li_wait_mult); 2252 if (nodefs || sbi->s_max_dir_size_kb) 2253 SEQ_OPTS_PRINT("max_dir_size_kb=%u", sbi->s_max_dir_size_kb); 2254 if (test_opt(sb, DATA_ERR_ABORT)) 2255 SEQ_OPTS_PUTS("data_err=abort"); 2256 if (DUMMY_ENCRYPTION_ENABLED(sbi)) 2257 SEQ_OPTS_PUTS("test_dummy_encryption"); 2258 2259 ext4_show_quota_options(seq, sb); 2260 return 0; 2261 } 2262 2263 static int ext4_show_options(struct seq_file *seq, struct dentry *root) 2264 { 2265 return _ext4_show_options(seq, root->d_sb, 0); 2266 } 2267 2268 int ext4_seq_options_show(struct seq_file *seq, void *offset) 2269 { 2270 struct super_block *sb = seq->private; 2271 int rc; 2272 2273 seq_puts(seq, sb_rdonly(sb) ? "ro" : "rw"); 2274 rc = _ext4_show_options(seq, sb, 1); 2275 seq_puts(seq, "\n"); 2276 return rc; 2277 } 2278 2279 static int ext4_setup_super(struct super_block *sb, struct ext4_super_block *es, 2280 int read_only) 2281 { 2282 struct ext4_sb_info *sbi = EXT4_SB(sb); 2283 int err = 0; 2284 2285 if (le32_to_cpu(es->s_rev_level) > EXT4_MAX_SUPP_REV) { 2286 ext4_msg(sb, KERN_ERR, "revision level too high, " 2287 "forcing read-only mode"); 2288 err = -EROFS; 2289 } 2290 if (read_only) 2291 goto done; 2292 if (!(sbi->s_mount_state & EXT4_VALID_FS)) 2293 ext4_msg(sb, KERN_WARNING, "warning: mounting unchecked fs, " 2294 "running e2fsck is recommended"); 2295 else if (sbi->s_mount_state & EXT4_ERROR_FS) 2296 ext4_msg(sb, KERN_WARNING, 2297 "warning: mounting fs with errors, " 2298 "running e2fsck is recommended"); 2299 else if ((__s16) le16_to_cpu(es->s_max_mnt_count) > 0 && 2300 le16_to_cpu(es->s_mnt_count) >= 2301 (unsigned short) (__s16) le16_to_cpu(es->s_max_mnt_count)) 2302 ext4_msg(sb, KERN_WARNING, 2303 "warning: maximal mount count reached, " 2304 "running e2fsck is recommended"); 2305 else if (le32_to_cpu(es->s_checkinterval) && 2306 (ext4_get_tstamp(es, s_lastcheck) + 2307 le32_to_cpu(es->s_checkinterval) <= ktime_get_real_seconds())) 2308 ext4_msg(sb, KERN_WARNING, 2309 "warning: checktime reached, " 2310 "running e2fsck is recommended"); 2311 if (!sbi->s_journal) 2312 es->s_state &= cpu_to_le16(~EXT4_VALID_FS); 2313 if (!(__s16) le16_to_cpu(es->s_max_mnt_count)) 2314 es->s_max_mnt_count = cpu_to_le16(EXT4_DFL_MAX_MNT_COUNT); 2315 le16_add_cpu(&es->s_mnt_count, 1); 2316 ext4_update_tstamp(es, s_mtime); 2317 if (sbi->s_journal) 2318 ext4_set_feature_journal_needs_recovery(sb); 2319 2320 err = ext4_commit_super(sb, 1); 2321 done: 2322 if (test_opt(sb, DEBUG)) 2323 printk(KERN_INFO "[EXT4 FS bs=%lu, gc=%u, " 2324 "bpg=%lu, ipg=%lu, mo=%04x, mo2=%04x]\n", 2325 sb->s_blocksize, 2326 sbi->s_groups_count, 2327 EXT4_BLOCKS_PER_GROUP(sb), 2328 EXT4_INODES_PER_GROUP(sb), 2329 sbi->s_mount_opt, sbi->s_mount_opt2); 2330 2331 cleancache_init_fs(sb); 2332 return err; 2333 } 2334 2335 int ext4_alloc_flex_bg_array(struct super_block *sb, ext4_group_t ngroup) 2336 { 2337 struct ext4_sb_info *sbi = EXT4_SB(sb); 2338 struct flex_groups *new_groups; 2339 int size; 2340 2341 if (!sbi->s_log_groups_per_flex) 2342 return 0; 2343 2344 size = ext4_flex_group(sbi, ngroup - 1) + 1; 2345 if (size <= sbi->s_flex_groups_allocated) 2346 return 0; 2347 2348 size = roundup_pow_of_two(size * sizeof(struct flex_groups)); 2349 new_groups = kvzalloc(size, GFP_KERNEL); 2350 if (!new_groups) { 2351 ext4_msg(sb, KERN_ERR, "not enough memory for %d flex groups", 2352 size / (int) sizeof(struct flex_groups)); 2353 return -ENOMEM; 2354 } 2355 2356 if (sbi->s_flex_groups) { 2357 memcpy(new_groups, sbi->s_flex_groups, 2358 (sbi->s_flex_groups_allocated * 2359 sizeof(struct flex_groups))); 2360 kvfree(sbi->s_flex_groups); 2361 } 2362 sbi->s_flex_groups = new_groups; 2363 sbi->s_flex_groups_allocated = size / sizeof(struct flex_groups); 2364 return 0; 2365 } 2366 2367 static int ext4_fill_flex_info(struct super_block *sb) 2368 { 2369 struct ext4_sb_info *sbi = EXT4_SB(sb); 2370 struct ext4_group_desc *gdp = NULL; 2371 ext4_group_t flex_group; 2372 int i, err; 2373 2374 sbi->s_log_groups_per_flex = sbi->s_es->s_log_groups_per_flex; 2375 if (sbi->s_log_groups_per_flex < 1 || sbi->s_log_groups_per_flex > 31) { 2376 sbi->s_log_groups_per_flex = 0; 2377 return 1; 2378 } 2379 2380 err = ext4_alloc_flex_bg_array(sb, sbi->s_groups_count); 2381 if (err) 2382 goto failed; 2383 2384 for (i = 0; i < sbi->s_groups_count; i++) { 2385 gdp = ext4_get_group_desc(sb, i, NULL); 2386 2387 flex_group = ext4_flex_group(sbi, i); 2388 atomic_add(ext4_free_inodes_count(sb, gdp), 2389 &sbi->s_flex_groups[flex_group].free_inodes); 2390 atomic64_add(ext4_free_group_clusters(sb, gdp), 2391 &sbi->s_flex_groups[flex_group].free_clusters); 2392 atomic_add(ext4_used_dirs_count(sb, gdp), 2393 &sbi->s_flex_groups[flex_group].used_dirs); 2394 } 2395 2396 return 1; 2397 failed: 2398 return 0; 2399 } 2400 2401 static __le16 ext4_group_desc_csum(struct super_block *sb, __u32 block_group, 2402 struct ext4_group_desc *gdp) 2403 { 2404 int offset = offsetof(struct ext4_group_desc, bg_checksum); 2405 __u16 crc = 0; 2406 __le32 le_group = cpu_to_le32(block_group); 2407 struct ext4_sb_info *sbi = EXT4_SB(sb); 2408 2409 if (ext4_has_metadata_csum(sbi->s_sb)) { 2410 /* Use new metadata_csum algorithm */ 2411 __u32 csum32; 2412 __u16 dummy_csum = 0; 2413 2414 csum32 = ext4_chksum(sbi, sbi->s_csum_seed, (__u8 *)&le_group, 2415 sizeof(le_group)); 2416 csum32 = ext4_chksum(sbi, csum32, (__u8 *)gdp, offset); 2417 csum32 = ext4_chksum(sbi, csum32, (__u8 *)&dummy_csum, 2418 sizeof(dummy_csum)); 2419 offset += sizeof(dummy_csum); 2420 if (offset < sbi->s_desc_size) 2421 csum32 = ext4_chksum(sbi, csum32, (__u8 *)gdp + offset, 2422 sbi->s_desc_size - offset); 2423 2424 crc = csum32 & 0xFFFF; 2425 goto out; 2426 } 2427 2428 /* old crc16 code */ 2429 if (!ext4_has_feature_gdt_csum(sb)) 2430 return 0; 2431 2432 crc = crc16(~0, sbi->s_es->s_uuid, sizeof(sbi->s_es->s_uuid)); 2433 crc = crc16(crc, (__u8 *)&le_group, sizeof(le_group)); 2434 crc = crc16(crc, (__u8 *)gdp, offset); 2435 offset += sizeof(gdp->bg_checksum); /* skip checksum */ 2436 /* for checksum of struct ext4_group_desc do the rest...*/ 2437 if (ext4_has_feature_64bit(sb) && 2438 offset < le16_to_cpu(sbi->s_es->s_desc_size)) 2439 crc = crc16(crc, (__u8 *)gdp + offset, 2440 le16_to_cpu(sbi->s_es->s_desc_size) - 2441 offset); 2442 2443 out: 2444 return cpu_to_le16(crc); 2445 } 2446 2447 int ext4_group_desc_csum_verify(struct super_block *sb, __u32 block_group, 2448 struct ext4_group_desc *gdp) 2449 { 2450 if (ext4_has_group_desc_csum(sb) && 2451 (gdp->bg_checksum != ext4_group_desc_csum(sb, block_group, gdp))) 2452 return 0; 2453 2454 return 1; 2455 } 2456 2457 void ext4_group_desc_csum_set(struct super_block *sb, __u32 block_group, 2458 struct ext4_group_desc *gdp) 2459 { 2460 if (!ext4_has_group_desc_csum(sb)) 2461 return; 2462 gdp->bg_checksum = ext4_group_desc_csum(sb, block_group, gdp); 2463 } 2464 2465 /* Called at mount-time, super-block is locked */ 2466 static int ext4_check_descriptors(struct super_block *sb, 2467 ext4_fsblk_t sb_block, 2468 ext4_group_t *first_not_zeroed) 2469 { 2470 struct ext4_sb_info *sbi = EXT4_SB(sb); 2471 ext4_fsblk_t first_block = le32_to_cpu(sbi->s_es->s_first_data_block); 2472 ext4_fsblk_t last_block; 2473 ext4_fsblk_t last_bg_block = sb_block + ext4_bg_num_gdb(sb, 0); 2474 ext4_fsblk_t block_bitmap; 2475 ext4_fsblk_t inode_bitmap; 2476 ext4_fsblk_t inode_table; 2477 int flexbg_flag = 0; 2478 ext4_group_t i, grp = sbi->s_groups_count; 2479 2480 if (ext4_has_feature_flex_bg(sb)) 2481 flexbg_flag = 1; 2482 2483 ext4_debug("Checking group descriptors"); 2484 2485 for (i = 0; i < sbi->s_groups_count; i++) { 2486 struct ext4_group_desc *gdp = ext4_get_group_desc(sb, i, NULL); 2487 2488 if (i == sbi->s_groups_count - 1 || flexbg_flag) 2489 last_block = ext4_blocks_count(sbi->s_es) - 1; 2490 else 2491 last_block = first_block + 2492 (EXT4_BLOCKS_PER_GROUP(sb) - 1); 2493 2494 if ((grp == sbi->s_groups_count) && 2495 !(gdp->bg_flags & cpu_to_le16(EXT4_BG_INODE_ZEROED))) 2496 grp = i; 2497 2498 block_bitmap = ext4_block_bitmap(sb, gdp); 2499 if (block_bitmap == sb_block) { 2500 ext4_msg(sb, KERN_ERR, "ext4_check_descriptors: " 2501 "Block bitmap for group %u overlaps " 2502 "superblock", i); 2503 if (!sb_rdonly(sb)) 2504 return 0; 2505 } 2506 if (block_bitmap >= sb_block + 1 && 2507 block_bitmap <= last_bg_block) { 2508 ext4_msg(sb, KERN_ERR, "ext4_check_descriptors: " 2509 "Block bitmap for group %u overlaps " 2510 "block group descriptors", i); 2511 if (!sb_rdonly(sb)) 2512 return 0; 2513 } 2514 if (block_bitmap < first_block || block_bitmap > last_block) { 2515 ext4_msg(sb, KERN_ERR, "ext4_check_descriptors: " 2516 "Block bitmap for group %u not in group " 2517 "(block %llu)!", i, block_bitmap); 2518 return 0; 2519 } 2520 inode_bitmap = ext4_inode_bitmap(sb, gdp); 2521 if (inode_bitmap == sb_block) { 2522 ext4_msg(sb, KERN_ERR, "ext4_check_descriptors: " 2523 "Inode bitmap for group %u overlaps " 2524 "superblock", i); 2525 if (!sb_rdonly(sb)) 2526 return 0; 2527 } 2528 if (inode_bitmap >= sb_block + 1 && 2529 inode_bitmap <= last_bg_block) { 2530 ext4_msg(sb, KERN_ERR, "ext4_check_descriptors: " 2531 "Inode bitmap for group %u overlaps " 2532 "block group descriptors", i); 2533 if (!sb_rdonly(sb)) 2534 return 0; 2535 } 2536 if (inode_bitmap < first_block || inode_bitmap > last_block) { 2537 ext4_msg(sb, KERN_ERR, "ext4_check_descriptors: " 2538 "Inode bitmap for group %u not in group " 2539 "(block %llu)!", i, inode_bitmap); 2540 return 0; 2541 } 2542 inode_table = ext4_inode_table(sb, gdp); 2543 if (inode_table == sb_block) { 2544 ext4_msg(sb, KERN_ERR, "ext4_check_descriptors: " 2545 "Inode table for group %u overlaps " 2546 "superblock", i); 2547 if (!sb_rdonly(sb)) 2548 return 0; 2549 } 2550 if (inode_table >= sb_block + 1 && 2551 inode_table <= last_bg_block) { 2552 ext4_msg(sb, KERN_ERR, "ext4_check_descriptors: " 2553 "Inode table for group %u overlaps " 2554 "block group descriptors", i); 2555 if (!sb_rdonly(sb)) 2556 return 0; 2557 } 2558 if (inode_table < first_block || 2559 inode_table + sbi->s_itb_per_group - 1 > last_block) { 2560 ext4_msg(sb, KERN_ERR, "ext4_check_descriptors: " 2561 "Inode table for group %u not in group " 2562 "(block %llu)!", i, inode_table); 2563 return 0; 2564 } 2565 ext4_lock_group(sb, i); 2566 if (!ext4_group_desc_csum_verify(sb, i, gdp)) { 2567 ext4_msg(sb, KERN_ERR, "ext4_check_descriptors: " 2568 "Checksum for group %u failed (%u!=%u)", 2569 i, le16_to_cpu(ext4_group_desc_csum(sb, i, 2570 gdp)), le16_to_cpu(gdp->bg_checksum)); 2571 if (!sb_rdonly(sb)) { 2572 ext4_unlock_group(sb, i); 2573 return 0; 2574 } 2575 } 2576 ext4_unlock_group(sb, i); 2577 if (!flexbg_flag) 2578 first_block += EXT4_BLOCKS_PER_GROUP(sb); 2579 } 2580 if (NULL != first_not_zeroed) 2581 *first_not_zeroed = grp; 2582 return 1; 2583 } 2584 2585 /* ext4_orphan_cleanup() walks a singly-linked list of inodes (starting at 2586 * the superblock) which were deleted from all directories, but held open by 2587 * a process at the time of a crash. We walk the list and try to delete these 2588 * inodes at recovery time (only with a read-write filesystem). 2589 * 2590 * In order to keep the orphan inode chain consistent during traversal (in 2591 * case of crash during recovery), we link each inode into the superblock 2592 * orphan list_head and handle it the same way as an inode deletion during 2593 * normal operation (which journals the operations for us). 2594 * 2595 * We only do an iget() and an iput() on each inode, which is very safe if we 2596 * accidentally point at an in-use or already deleted inode. The worst that 2597 * can happen in this case is that we get a "bit already cleared" message from 2598 * ext4_free_inode(). The only reason we would point at a wrong inode is if 2599 * e2fsck was run on this filesystem, and it must have already done the orphan 2600 * inode cleanup for us, so we can safely abort without any further action. 2601 */ 2602 static void ext4_orphan_cleanup(struct super_block *sb, 2603 struct ext4_super_block *es) 2604 { 2605 unsigned int s_flags = sb->s_flags; 2606 int ret, nr_orphans = 0, nr_truncates = 0; 2607 #ifdef CONFIG_QUOTA 2608 int quota_update = 0; 2609 int i; 2610 #endif 2611 if (!es->s_last_orphan) { 2612 jbd_debug(4, "no orphan inodes to clean up\n"); 2613 return; 2614 } 2615 2616 if (bdev_read_only(sb->s_bdev)) { 2617 ext4_msg(sb, KERN_ERR, "write access " 2618 "unavailable, skipping orphan cleanup"); 2619 return; 2620 } 2621 2622 /* Check if feature set would not allow a r/w mount */ 2623 if (!ext4_feature_set_ok(sb, 0)) { 2624 ext4_msg(sb, KERN_INFO, "Skipping orphan cleanup due to " 2625 "unknown ROCOMPAT features"); 2626 return; 2627 } 2628 2629 if (EXT4_SB(sb)->s_mount_state & EXT4_ERROR_FS) { 2630 /* don't clear list on RO mount w/ errors */ 2631 if (es->s_last_orphan && !(s_flags & SB_RDONLY)) { 2632 ext4_msg(sb, KERN_INFO, "Errors on filesystem, " 2633 "clearing orphan list.\n"); 2634 es->s_last_orphan = 0; 2635 } 2636 jbd_debug(1, "Skipping orphan recovery on fs with errors.\n"); 2637 return; 2638 } 2639 2640 if (s_flags & SB_RDONLY) { 2641 ext4_msg(sb, KERN_INFO, "orphan cleanup on readonly fs"); 2642 sb->s_flags &= ~SB_RDONLY; 2643 } 2644 #ifdef CONFIG_QUOTA 2645 /* Needed for iput() to work correctly and not trash data */ 2646 sb->s_flags |= SB_ACTIVE; 2647 2648 /* 2649 * Turn on quotas which were not enabled for read-only mounts if 2650 * filesystem has quota feature, so that they are updated correctly. 2651 */ 2652 if (ext4_has_feature_quota(sb) && (s_flags & SB_RDONLY)) { 2653 int ret = ext4_enable_quotas(sb); 2654 2655 if (!ret) 2656 quota_update = 1; 2657 else 2658 ext4_msg(sb, KERN_ERR, 2659 "Cannot turn on quotas: error %d", ret); 2660 } 2661 2662 /* Turn on journaled quotas used for old sytle */ 2663 for (i = 0; i < EXT4_MAXQUOTAS; i++) { 2664 if (EXT4_SB(sb)->s_qf_names[i]) { 2665 int ret = ext4_quota_on_mount(sb, i); 2666 2667 if (!ret) 2668 quota_update = 1; 2669 else 2670 ext4_msg(sb, KERN_ERR, 2671 "Cannot turn on journaled " 2672 "quota: type %d: error %d", i, ret); 2673 } 2674 } 2675 #endif 2676 2677 while (es->s_last_orphan) { 2678 struct inode *inode; 2679 2680 /* 2681 * We may have encountered an error during cleanup; if 2682 * so, skip the rest. 2683 */ 2684 if (EXT4_SB(sb)->s_mount_state & EXT4_ERROR_FS) { 2685 jbd_debug(1, "Skipping orphan recovery on fs with errors.\n"); 2686 es->s_last_orphan = 0; 2687 break; 2688 } 2689 2690 inode = ext4_orphan_get(sb, le32_to_cpu(es->s_last_orphan)); 2691 if (IS_ERR(inode)) { 2692 es->s_last_orphan = 0; 2693 break; 2694 } 2695 2696 list_add(&EXT4_I(inode)->i_orphan, &EXT4_SB(sb)->s_orphan); 2697 dquot_initialize(inode); 2698 if (inode->i_nlink) { 2699 if (test_opt(sb, DEBUG)) 2700 ext4_msg(sb, KERN_DEBUG, 2701 "%s: truncating inode %lu to %lld bytes", 2702 __func__, inode->i_ino, inode->i_size); 2703 jbd_debug(2, "truncating inode %lu to %lld bytes\n", 2704 inode->i_ino, inode->i_size); 2705 inode_lock(inode); 2706 truncate_inode_pages(inode->i_mapping, inode->i_size); 2707 ret = ext4_truncate(inode); 2708 if (ret) 2709 ext4_std_error(inode->i_sb, ret); 2710 inode_unlock(inode); 2711 nr_truncates++; 2712 } else { 2713 if (test_opt(sb, DEBUG)) 2714 ext4_msg(sb, KERN_DEBUG, 2715 "%s: deleting unreferenced inode %lu", 2716 __func__, inode->i_ino); 2717 jbd_debug(2, "deleting unreferenced inode %lu\n", 2718 inode->i_ino); 2719 nr_orphans++; 2720 } 2721 iput(inode); /* The delete magic happens here! */ 2722 } 2723 2724 #define PLURAL(x) (x), ((x) == 1) ? "" : "s" 2725 2726 if (nr_orphans) 2727 ext4_msg(sb, KERN_INFO, "%d orphan inode%s deleted", 2728 PLURAL(nr_orphans)); 2729 if (nr_truncates) 2730 ext4_msg(sb, KERN_INFO, "%d truncate%s cleaned up", 2731 PLURAL(nr_truncates)); 2732 #ifdef CONFIG_QUOTA 2733 /* Turn off quotas if they were enabled for orphan cleanup */ 2734 if (quota_update) { 2735 for (i = 0; i < EXT4_MAXQUOTAS; i++) { 2736 if (sb_dqopt(sb)->files[i]) 2737 dquot_quota_off(sb, i); 2738 } 2739 } 2740 #endif 2741 sb->s_flags = s_flags; /* Restore SB_RDONLY status */ 2742 } 2743 2744 /* 2745 * Maximal extent format file size. 2746 * Resulting logical blkno at s_maxbytes must fit in our on-disk 2747 * extent format containers, within a sector_t, and within i_blocks 2748 * in the vfs. ext4 inode has 48 bits of i_block in fsblock units, 2749 * so that won't be a limiting factor. 2750 * 2751 * However there is other limiting factor. We do store extents in the form 2752 * of starting block and length, hence the resulting length of the extent 2753 * covering maximum file size must fit into on-disk format containers as 2754 * well. Given that length is always by 1 unit bigger than max unit (because 2755 * we count 0 as well) we have to lower the s_maxbytes by one fs block. 2756 * 2757 * Note, this does *not* consider any metadata overhead for vfs i_blocks. 2758 */ 2759 static loff_t ext4_max_size(int blkbits, int has_huge_files) 2760 { 2761 loff_t res; 2762 loff_t upper_limit = MAX_LFS_FILESIZE; 2763 2764 BUILD_BUG_ON(sizeof(blkcnt_t) < sizeof(u64)); 2765 2766 if (!has_huge_files) { 2767 upper_limit = (1LL << 32) - 1; 2768 2769 /* total blocks in file system block size */ 2770 upper_limit >>= (blkbits - 9); 2771 upper_limit <<= blkbits; 2772 } 2773 2774 /* 2775 * 32-bit extent-start container, ee_block. We lower the maxbytes 2776 * by one fs block, so ee_len can cover the extent of maximum file 2777 * size 2778 */ 2779 res = (1LL << 32) - 1; 2780 res <<= blkbits; 2781 2782 /* Sanity check against vm- & vfs- imposed limits */ 2783 if (res > upper_limit) 2784 res = upper_limit; 2785 2786 return res; 2787 } 2788 2789 /* 2790 * Maximal bitmap file size. There is a direct, and {,double-,triple-}indirect 2791 * block limit, and also a limit of (2^48 - 1) 512-byte sectors in i_blocks. 2792 * We need to be 1 filesystem block less than the 2^48 sector limit. 2793 */ 2794 static loff_t ext4_max_bitmap_size(int bits, int has_huge_files) 2795 { 2796 loff_t res = EXT4_NDIR_BLOCKS; 2797 int meta_blocks; 2798 loff_t upper_limit; 2799 /* This is calculated to be the largest file size for a dense, block 2800 * mapped file such that the file's total number of 512-byte sectors, 2801 * including data and all indirect blocks, does not exceed (2^48 - 1). 2802 * 2803 * __u32 i_blocks_lo and _u16 i_blocks_high represent the total 2804 * number of 512-byte sectors of the file. 2805 */ 2806 2807 if (!has_huge_files) { 2808 /* 2809 * !has_huge_files or implies that the inode i_block field 2810 * represents total file blocks in 2^32 512-byte sectors == 2811 * size of vfs inode i_blocks * 8 2812 */ 2813 upper_limit = (1LL << 32) - 1; 2814 2815 /* total blocks in file system block size */ 2816 upper_limit >>= (bits - 9); 2817 2818 } else { 2819 /* 2820 * We use 48 bit ext4_inode i_blocks 2821 * With EXT4_HUGE_FILE_FL set the i_blocks 2822 * represent total number of blocks in 2823 * file system block size 2824 */ 2825 upper_limit = (1LL << 48) - 1; 2826 2827 } 2828 2829 /* indirect blocks */ 2830 meta_blocks = 1; 2831 /* double indirect blocks */ 2832 meta_blocks += 1 + (1LL << (bits-2)); 2833 /* tripple indirect blocks */ 2834 meta_blocks += 1 + (1LL << (bits-2)) + (1LL << (2*(bits-2))); 2835 2836 upper_limit -= meta_blocks; 2837 upper_limit <<= bits; 2838 2839 res += 1LL << (bits-2); 2840 res += 1LL << (2*(bits-2)); 2841 res += 1LL << (3*(bits-2)); 2842 res <<= bits; 2843 if (res > upper_limit) 2844 res = upper_limit; 2845 2846 if (res > MAX_LFS_FILESIZE) 2847 res = MAX_LFS_FILESIZE; 2848 2849 return res; 2850 } 2851 2852 static ext4_fsblk_t descriptor_loc(struct super_block *sb, 2853 ext4_fsblk_t logical_sb_block, int nr) 2854 { 2855 struct ext4_sb_info *sbi = EXT4_SB(sb); 2856 ext4_group_t bg, first_meta_bg; 2857 int has_super = 0; 2858 2859 first_meta_bg = le32_to_cpu(sbi->s_es->s_first_meta_bg); 2860 2861 if (!ext4_has_feature_meta_bg(sb) || nr < first_meta_bg) 2862 return logical_sb_block + nr + 1; 2863 bg = sbi->s_desc_per_block * nr; 2864 if (ext4_bg_has_super(sb, bg)) 2865 has_super = 1; 2866 2867 /* 2868 * If we have a meta_bg fs with 1k blocks, group 0's GDT is at 2869 * block 2, not 1. If s_first_data_block == 0 (bigalloc is enabled 2870 * on modern mke2fs or blksize > 1k on older mke2fs) then we must 2871 * compensate. 2872 */ 2873 if (sb->s_blocksize == 1024 && nr == 0 && 2874 le32_to_cpu(sbi->s_es->s_first_data_block) == 0) 2875 has_super++; 2876 2877 return (has_super + ext4_group_first_block_no(sb, bg)); 2878 } 2879 2880 /** 2881 * ext4_get_stripe_size: Get the stripe size. 2882 * @sbi: In memory super block info 2883 * 2884 * If we have specified it via mount option, then 2885 * use the mount option value. If the value specified at mount time is 2886 * greater than the blocks per group use the super block value. 2887 * If the super block value is greater than blocks per group return 0. 2888 * Allocator needs it be less than blocks per group. 2889 * 2890 */ 2891 static unsigned long ext4_get_stripe_size(struct ext4_sb_info *sbi) 2892 { 2893 unsigned long stride = le16_to_cpu(sbi->s_es->s_raid_stride); 2894 unsigned long stripe_width = 2895 le32_to_cpu(sbi->s_es->s_raid_stripe_width); 2896 int ret; 2897 2898 if (sbi->s_stripe && sbi->s_stripe <= sbi->s_blocks_per_group) 2899 ret = sbi->s_stripe; 2900 else if (stripe_width && stripe_width <= sbi->s_blocks_per_group) 2901 ret = stripe_width; 2902 else if (stride && stride <= sbi->s_blocks_per_group) 2903 ret = stride; 2904 else 2905 ret = 0; 2906 2907 /* 2908 * If the stripe width is 1, this makes no sense and 2909 * we set it to 0 to turn off stripe handling code. 2910 */ 2911 if (ret <= 1) 2912 ret = 0; 2913 2914 return ret; 2915 } 2916 2917 /* 2918 * Check whether this filesystem can be mounted based on 2919 * the features present and the RDONLY/RDWR mount requested. 2920 * Returns 1 if this filesystem can be mounted as requested, 2921 * 0 if it cannot be. 2922 */ 2923 static int ext4_feature_set_ok(struct super_block *sb, int readonly) 2924 { 2925 if (ext4_has_unknown_ext4_incompat_features(sb)) { 2926 ext4_msg(sb, KERN_ERR, 2927 "Couldn't mount because of " 2928 "unsupported optional features (%x)", 2929 (le32_to_cpu(EXT4_SB(sb)->s_es->s_feature_incompat) & 2930 ~EXT4_FEATURE_INCOMPAT_SUPP)); 2931 return 0; 2932 } 2933 2934 #ifndef CONFIG_UNICODE 2935 if (ext4_has_feature_casefold(sb)) { 2936 ext4_msg(sb, KERN_ERR, 2937 "Filesystem with casefold feature cannot be " 2938 "mounted without CONFIG_UNICODE"); 2939 return 0; 2940 } 2941 #endif 2942 2943 if (readonly) 2944 return 1; 2945 2946 if (ext4_has_feature_readonly(sb)) { 2947 ext4_msg(sb, KERN_INFO, "filesystem is read-only"); 2948 sb->s_flags |= SB_RDONLY; 2949 return 1; 2950 } 2951 2952 /* Check that feature set is OK for a read-write mount */ 2953 if (ext4_has_unknown_ext4_ro_compat_features(sb)) { 2954 ext4_msg(sb, KERN_ERR, "couldn't mount RDWR because of " 2955 "unsupported optional features (%x)", 2956 (le32_to_cpu(EXT4_SB(sb)->s_es->s_feature_ro_compat) & 2957 ~EXT4_FEATURE_RO_COMPAT_SUPP)); 2958 return 0; 2959 } 2960 if (ext4_has_feature_bigalloc(sb) && !ext4_has_feature_extents(sb)) { 2961 ext4_msg(sb, KERN_ERR, 2962 "Can't support bigalloc feature without " 2963 "extents feature\n"); 2964 return 0; 2965 } 2966 2967 #ifndef CONFIG_QUOTA 2968 if (ext4_has_feature_quota(sb) && !readonly) { 2969 ext4_msg(sb, KERN_ERR, 2970 "Filesystem with quota feature cannot be mounted RDWR " 2971 "without CONFIG_QUOTA"); 2972 return 0; 2973 } 2974 if (ext4_has_feature_project(sb) && !readonly) { 2975 ext4_msg(sb, KERN_ERR, 2976 "Filesystem with project quota feature cannot be mounted RDWR " 2977 "without CONFIG_QUOTA"); 2978 return 0; 2979 } 2980 #endif /* CONFIG_QUOTA */ 2981 return 1; 2982 } 2983 2984 /* 2985 * This function is called once a day if we have errors logged 2986 * on the file system 2987 */ 2988 static void print_daily_error_info(struct timer_list *t) 2989 { 2990 struct ext4_sb_info *sbi = from_timer(sbi, t, s_err_report); 2991 struct super_block *sb = sbi->s_sb; 2992 struct ext4_super_block *es = sbi->s_es; 2993 2994 if (es->s_error_count) 2995 /* fsck newer than v1.41.13 is needed to clean this condition. */ 2996 ext4_msg(sb, KERN_NOTICE, "error count since last fsck: %u", 2997 le32_to_cpu(es->s_error_count)); 2998 if (es->s_first_error_time) { 2999 printk(KERN_NOTICE "EXT4-fs (%s): initial error at time %llu: %.*s:%d", 3000 sb->s_id, 3001 ext4_get_tstamp(es, s_first_error_time), 3002 (int) sizeof(es->s_first_error_func), 3003 es->s_first_error_func, 3004 le32_to_cpu(es->s_first_error_line)); 3005 if (es->s_first_error_ino) 3006 printk(KERN_CONT ": inode %u", 3007 le32_to_cpu(es->s_first_error_ino)); 3008 if (es->s_first_error_block) 3009 printk(KERN_CONT ": block %llu", (unsigned long long) 3010 le64_to_cpu(es->s_first_error_block)); 3011 printk(KERN_CONT "\n"); 3012 } 3013 if (es->s_last_error_time) { 3014 printk(KERN_NOTICE "EXT4-fs (%s): last error at time %llu: %.*s:%d", 3015 sb->s_id, 3016 ext4_get_tstamp(es, s_last_error_time), 3017 (int) sizeof(es->s_last_error_func), 3018 es->s_last_error_func, 3019 le32_to_cpu(es->s_last_error_line)); 3020 if (es->s_last_error_ino) 3021 printk(KERN_CONT ": inode %u", 3022 le32_to_cpu(es->s_last_error_ino)); 3023 if (es->s_last_error_block) 3024 printk(KERN_CONT ": block %llu", (unsigned long long) 3025 le64_to_cpu(es->s_last_error_block)); 3026 printk(KERN_CONT "\n"); 3027 } 3028 mod_timer(&sbi->s_err_report, jiffies + 24*60*60*HZ); /* Once a day */ 3029 } 3030 3031 /* Find next suitable group and run ext4_init_inode_table */ 3032 static int ext4_run_li_request(struct ext4_li_request *elr) 3033 { 3034 struct ext4_group_desc *gdp = NULL; 3035 ext4_group_t group, ngroups; 3036 struct super_block *sb; 3037 unsigned long timeout = 0; 3038 int ret = 0; 3039 3040 sb = elr->lr_super; 3041 ngroups = EXT4_SB(sb)->s_groups_count; 3042 3043 for (group = elr->lr_next_group; group < ngroups; group++) { 3044 gdp = ext4_get_group_desc(sb, group, NULL); 3045 if (!gdp) { 3046 ret = 1; 3047 break; 3048 } 3049 3050 if (!(gdp->bg_flags & cpu_to_le16(EXT4_BG_INODE_ZEROED))) 3051 break; 3052 } 3053 3054 if (group >= ngroups) 3055 ret = 1; 3056 3057 if (!ret) { 3058 timeout = jiffies; 3059 ret = ext4_init_inode_table(sb, group, 3060 elr->lr_timeout ? 0 : 1); 3061 if (elr->lr_timeout == 0) { 3062 timeout = (jiffies - timeout) * 3063 elr->lr_sbi->s_li_wait_mult; 3064 elr->lr_timeout = timeout; 3065 } 3066 elr->lr_next_sched = jiffies + elr->lr_timeout; 3067 elr->lr_next_group = group + 1; 3068 } 3069 return ret; 3070 } 3071 3072 /* 3073 * Remove lr_request from the list_request and free the 3074 * request structure. Should be called with li_list_mtx held 3075 */ 3076 static void ext4_remove_li_request(struct ext4_li_request *elr) 3077 { 3078 struct ext4_sb_info *sbi; 3079 3080 if (!elr) 3081 return; 3082 3083 sbi = elr->lr_sbi; 3084 3085 list_del(&elr->lr_request); 3086 sbi->s_li_request = NULL; 3087 kfree(elr); 3088 } 3089 3090 static void ext4_unregister_li_request(struct super_block *sb) 3091 { 3092 mutex_lock(&ext4_li_mtx); 3093 if (!ext4_li_info) { 3094 mutex_unlock(&ext4_li_mtx); 3095 return; 3096 } 3097 3098 mutex_lock(&ext4_li_info->li_list_mtx); 3099 ext4_remove_li_request(EXT4_SB(sb)->s_li_request); 3100 mutex_unlock(&ext4_li_info->li_list_mtx); 3101 mutex_unlock(&ext4_li_mtx); 3102 } 3103 3104 static struct task_struct *ext4_lazyinit_task; 3105 3106 /* 3107 * This is the function where ext4lazyinit thread lives. It walks 3108 * through the request list searching for next scheduled filesystem. 3109 * When such a fs is found, run the lazy initialization request 3110 * (ext4_rn_li_request) and keep track of the time spend in this 3111 * function. Based on that time we compute next schedule time of 3112 * the request. When walking through the list is complete, compute 3113 * next waking time and put itself into sleep. 3114 */ 3115 static int ext4_lazyinit_thread(void *arg) 3116 { 3117 struct ext4_lazy_init *eli = (struct ext4_lazy_init *)arg; 3118 struct list_head *pos, *n; 3119 struct ext4_li_request *elr; 3120 unsigned long next_wakeup, cur; 3121 3122 BUG_ON(NULL == eli); 3123 3124 cont_thread: 3125 while (true) { 3126 next_wakeup = MAX_JIFFY_OFFSET; 3127 3128 mutex_lock(&eli->li_list_mtx); 3129 if (list_empty(&eli->li_request_list)) { 3130 mutex_unlock(&eli->li_list_mtx); 3131 goto exit_thread; 3132 } 3133 list_for_each_safe(pos, n, &eli->li_request_list) { 3134 int err = 0; 3135 int progress = 0; 3136 elr = list_entry(pos, struct ext4_li_request, 3137 lr_request); 3138 3139 if (time_before(jiffies, elr->lr_next_sched)) { 3140 if (time_before(elr->lr_next_sched, next_wakeup)) 3141 next_wakeup = elr->lr_next_sched; 3142 continue; 3143 } 3144 if (down_read_trylock(&elr->lr_super->s_umount)) { 3145 if (sb_start_write_trylock(elr->lr_super)) { 3146 progress = 1; 3147 /* 3148 * We hold sb->s_umount, sb can not 3149 * be removed from the list, it is 3150 * now safe to drop li_list_mtx 3151 */ 3152 mutex_unlock(&eli->li_list_mtx); 3153 err = ext4_run_li_request(elr); 3154 sb_end_write(elr->lr_super); 3155 mutex_lock(&eli->li_list_mtx); 3156 n = pos->next; 3157 } 3158 up_read((&elr->lr_super->s_umount)); 3159 } 3160 /* error, remove the lazy_init job */ 3161 if (err) { 3162 ext4_remove_li_request(elr); 3163 continue; 3164 } 3165 if (!progress) { 3166 elr->lr_next_sched = jiffies + 3167 (prandom_u32() 3168 % (EXT4_DEF_LI_MAX_START_DELAY * HZ)); 3169 } 3170 if (time_before(elr->lr_next_sched, next_wakeup)) 3171 next_wakeup = elr->lr_next_sched; 3172 } 3173 mutex_unlock(&eli->li_list_mtx); 3174 3175 try_to_freeze(); 3176 3177 cur = jiffies; 3178 if ((time_after_eq(cur, next_wakeup)) || 3179 (MAX_JIFFY_OFFSET == next_wakeup)) { 3180 cond_resched(); 3181 continue; 3182 } 3183 3184 schedule_timeout_interruptible(next_wakeup - cur); 3185 3186 if (kthread_should_stop()) { 3187 ext4_clear_request_list(); 3188 goto exit_thread; 3189 } 3190 } 3191 3192 exit_thread: 3193 /* 3194 * It looks like the request list is empty, but we need 3195 * to check it under the li_list_mtx lock, to prevent any 3196 * additions into it, and of course we should lock ext4_li_mtx 3197 * to atomically free the list and ext4_li_info, because at 3198 * this point another ext4 filesystem could be registering 3199 * new one. 3200 */ 3201 mutex_lock(&ext4_li_mtx); 3202 mutex_lock(&eli->li_list_mtx); 3203 if (!list_empty(&eli->li_request_list)) { 3204 mutex_unlock(&eli->li_list_mtx); 3205 mutex_unlock(&ext4_li_mtx); 3206 goto cont_thread; 3207 } 3208 mutex_unlock(&eli->li_list_mtx); 3209 kfree(ext4_li_info); 3210 ext4_li_info = NULL; 3211 mutex_unlock(&ext4_li_mtx); 3212 3213 return 0; 3214 } 3215 3216 static void ext4_clear_request_list(void) 3217 { 3218 struct list_head *pos, *n; 3219 struct ext4_li_request *elr; 3220 3221 mutex_lock(&ext4_li_info->li_list_mtx); 3222 list_for_each_safe(pos, n, &ext4_li_info->li_request_list) { 3223 elr = list_entry(pos, struct ext4_li_request, 3224 lr_request); 3225 ext4_remove_li_request(elr); 3226 } 3227 mutex_unlock(&ext4_li_info->li_list_mtx); 3228 } 3229 3230 static int ext4_run_lazyinit_thread(void) 3231 { 3232 ext4_lazyinit_task = kthread_run(ext4_lazyinit_thread, 3233 ext4_li_info, "ext4lazyinit"); 3234 if (IS_ERR(ext4_lazyinit_task)) { 3235 int err = PTR_ERR(ext4_lazyinit_task); 3236 ext4_clear_request_list(); 3237 kfree(ext4_li_info); 3238 ext4_li_info = NULL; 3239 printk(KERN_CRIT "EXT4-fs: error %d creating inode table " 3240 "initialization thread\n", 3241 err); 3242 return err; 3243 } 3244 ext4_li_info->li_state |= EXT4_LAZYINIT_RUNNING; 3245 return 0; 3246 } 3247 3248 /* 3249 * Check whether it make sense to run itable init. thread or not. 3250 * If there is at least one uninitialized inode table, return 3251 * corresponding group number, else the loop goes through all 3252 * groups and return total number of groups. 3253 */ 3254 static ext4_group_t ext4_has_uninit_itable(struct super_block *sb) 3255 { 3256 ext4_group_t group, ngroups = EXT4_SB(sb)->s_groups_count; 3257 struct ext4_group_desc *gdp = NULL; 3258 3259 if (!ext4_has_group_desc_csum(sb)) 3260 return ngroups; 3261 3262 for (group = 0; group < ngroups; group++) { 3263 gdp = ext4_get_group_desc(sb, group, NULL); 3264 if (!gdp) 3265 continue; 3266 3267 if (!(gdp->bg_flags & cpu_to_le16(EXT4_BG_INODE_ZEROED))) 3268 break; 3269 } 3270 3271 return group; 3272 } 3273 3274 static int ext4_li_info_new(void) 3275 { 3276 struct ext4_lazy_init *eli = NULL; 3277 3278 eli = kzalloc(sizeof(*eli), GFP_KERNEL); 3279 if (!eli) 3280 return -ENOMEM; 3281 3282 INIT_LIST_HEAD(&eli->li_request_list); 3283 mutex_init(&eli->li_list_mtx); 3284 3285 eli->li_state |= EXT4_LAZYINIT_QUIT; 3286 3287 ext4_li_info = eli; 3288 3289 return 0; 3290 } 3291 3292 static struct ext4_li_request *ext4_li_request_new(struct super_block *sb, 3293 ext4_group_t start) 3294 { 3295 struct ext4_sb_info *sbi = EXT4_SB(sb); 3296 struct ext4_li_request *elr; 3297 3298 elr = kzalloc(sizeof(*elr), GFP_KERNEL); 3299 if (!elr) 3300 return NULL; 3301 3302 elr->lr_super = sb; 3303 elr->lr_sbi = sbi; 3304 elr->lr_next_group = start; 3305 3306 /* 3307 * Randomize first schedule time of the request to 3308 * spread the inode table initialization requests 3309 * better. 3310 */ 3311 elr->lr_next_sched = jiffies + (prandom_u32() % 3312 (EXT4_DEF_LI_MAX_START_DELAY * HZ)); 3313 return elr; 3314 } 3315 3316 int ext4_register_li_request(struct super_block *sb, 3317 ext4_group_t first_not_zeroed) 3318 { 3319 struct ext4_sb_info *sbi = EXT4_SB(sb); 3320 struct ext4_li_request *elr = NULL; 3321 ext4_group_t ngroups = sbi->s_groups_count; 3322 int ret = 0; 3323 3324 mutex_lock(&ext4_li_mtx); 3325 if (sbi->s_li_request != NULL) { 3326 /* 3327 * Reset timeout so it can be computed again, because 3328 * s_li_wait_mult might have changed. 3329 */ 3330 sbi->s_li_request->lr_timeout = 0; 3331 goto out; 3332 } 3333 3334 if (first_not_zeroed == ngroups || sb_rdonly(sb) || 3335 !test_opt(sb, INIT_INODE_TABLE)) 3336 goto out; 3337 3338 elr = ext4_li_request_new(sb, first_not_zeroed); 3339 if (!elr) { 3340 ret = -ENOMEM; 3341 goto out; 3342 } 3343 3344 if (NULL == ext4_li_info) { 3345 ret = ext4_li_info_new(); 3346 if (ret) 3347 goto out; 3348 } 3349 3350 mutex_lock(&ext4_li_info->li_list_mtx); 3351 list_add(&elr->lr_request, &ext4_li_info->li_request_list); 3352 mutex_unlock(&ext4_li_info->li_list_mtx); 3353 3354 sbi->s_li_request = elr; 3355 /* 3356 * set elr to NULL here since it has been inserted to 3357 * the request_list and the removal and free of it is 3358 * handled by ext4_clear_request_list from now on. 3359 */ 3360 elr = NULL; 3361 3362 if (!(ext4_li_info->li_state & EXT4_LAZYINIT_RUNNING)) { 3363 ret = ext4_run_lazyinit_thread(); 3364 if (ret) 3365 goto out; 3366 } 3367 out: 3368 mutex_unlock(&ext4_li_mtx); 3369 if (ret) 3370 kfree(elr); 3371 return ret; 3372 } 3373 3374 /* 3375 * We do not need to lock anything since this is called on 3376 * module unload. 3377 */ 3378 static void ext4_destroy_lazyinit_thread(void) 3379 { 3380 /* 3381 * If thread exited earlier 3382 * there's nothing to be done. 3383 */ 3384 if (!ext4_li_info || !ext4_lazyinit_task) 3385 return; 3386 3387 kthread_stop(ext4_lazyinit_task); 3388 } 3389 3390 static int set_journal_csum_feature_set(struct super_block *sb) 3391 { 3392 int ret = 1; 3393 int compat, incompat; 3394 struct ext4_sb_info *sbi = EXT4_SB(sb); 3395 3396 if (ext4_has_metadata_csum(sb)) { 3397 /* journal checksum v3 */ 3398 compat = 0; 3399 incompat = JBD2_FEATURE_INCOMPAT_CSUM_V3; 3400 } else { 3401 /* journal checksum v1 */ 3402 compat = JBD2_FEATURE_COMPAT_CHECKSUM; 3403 incompat = 0; 3404 } 3405 3406 jbd2_journal_clear_features(sbi->s_journal, 3407 JBD2_FEATURE_COMPAT_CHECKSUM, 0, 3408 JBD2_FEATURE_INCOMPAT_CSUM_V3 | 3409 JBD2_FEATURE_INCOMPAT_CSUM_V2); 3410 if (test_opt(sb, JOURNAL_ASYNC_COMMIT)) { 3411 ret = jbd2_journal_set_features(sbi->s_journal, 3412 compat, 0, 3413 JBD2_FEATURE_INCOMPAT_ASYNC_COMMIT | 3414 incompat); 3415 } else if (test_opt(sb, JOURNAL_CHECKSUM)) { 3416 ret = jbd2_journal_set_features(sbi->s_journal, 3417 compat, 0, 3418 incompat); 3419 jbd2_journal_clear_features(sbi->s_journal, 0, 0, 3420 JBD2_FEATURE_INCOMPAT_ASYNC_COMMIT); 3421 } else { 3422 jbd2_journal_clear_features(sbi->s_journal, 0, 0, 3423 JBD2_FEATURE_INCOMPAT_ASYNC_COMMIT); 3424 } 3425 3426 return ret; 3427 } 3428 3429 /* 3430 * Note: calculating the overhead so we can be compatible with 3431 * historical BSD practice is quite difficult in the face of 3432 * clusters/bigalloc. This is because multiple metadata blocks from 3433 * different block group can end up in the same allocation cluster. 3434 * Calculating the exact overhead in the face of clustered allocation 3435 * requires either O(all block bitmaps) in memory or O(number of block 3436 * groups**2) in time. We will still calculate the superblock for 3437 * older file systems --- and if we come across with a bigalloc file 3438 * system with zero in s_overhead_clusters the estimate will be close to 3439 * correct especially for very large cluster sizes --- but for newer 3440 * file systems, it's better to calculate this figure once at mkfs 3441 * time, and store it in the superblock. If the superblock value is 3442 * present (even for non-bigalloc file systems), we will use it. 3443 */ 3444 static int count_overhead(struct super_block *sb, ext4_group_t grp, 3445 char *buf) 3446 { 3447 struct ext4_sb_info *sbi = EXT4_SB(sb); 3448 struct ext4_group_desc *gdp; 3449 ext4_fsblk_t first_block, last_block, b; 3450 ext4_group_t i, ngroups = ext4_get_groups_count(sb); 3451 int s, j, count = 0; 3452 3453 if (!ext4_has_feature_bigalloc(sb)) 3454 return (ext4_bg_has_super(sb, grp) + ext4_bg_num_gdb(sb, grp) + 3455 sbi->s_itb_per_group + 2); 3456 3457 first_block = le32_to_cpu(sbi->s_es->s_first_data_block) + 3458 (grp * EXT4_BLOCKS_PER_GROUP(sb)); 3459 last_block = first_block + EXT4_BLOCKS_PER_GROUP(sb) - 1; 3460 for (i = 0; i < ngroups; i++) { 3461 gdp = ext4_get_group_desc(sb, i, NULL); 3462 b = ext4_block_bitmap(sb, gdp); 3463 if (b >= first_block && b <= last_block) { 3464 ext4_set_bit(EXT4_B2C(sbi, b - first_block), buf); 3465 count++; 3466 } 3467 b = ext4_inode_bitmap(sb, gdp); 3468 if (b >= first_block && b <= last_block) { 3469 ext4_set_bit(EXT4_B2C(sbi, b - first_block), buf); 3470 count++; 3471 } 3472 b = ext4_inode_table(sb, gdp); 3473 if (b >= first_block && b + sbi->s_itb_per_group <= last_block) 3474 for (j = 0; j < sbi->s_itb_per_group; j++, b++) { 3475 int c = EXT4_B2C(sbi, b - first_block); 3476 ext4_set_bit(c, buf); 3477 count++; 3478 } 3479 if (i != grp) 3480 continue; 3481 s = 0; 3482 if (ext4_bg_has_super(sb, grp)) { 3483 ext4_set_bit(s++, buf); 3484 count++; 3485 } 3486 j = ext4_bg_num_gdb(sb, grp); 3487 if (s + j > EXT4_BLOCKS_PER_GROUP(sb)) { 3488 ext4_error(sb, "Invalid number of block group " 3489 "descriptor blocks: %d", j); 3490 j = EXT4_BLOCKS_PER_GROUP(sb) - s; 3491 } 3492 count += j; 3493 for (; j > 0; j--) 3494 ext4_set_bit(EXT4_B2C(sbi, s++), buf); 3495 } 3496 if (!count) 3497 return 0; 3498 return EXT4_CLUSTERS_PER_GROUP(sb) - 3499 ext4_count_free(buf, EXT4_CLUSTERS_PER_GROUP(sb) / 8); 3500 } 3501 3502 /* 3503 * Compute the overhead and stash it in sbi->s_overhead 3504 */ 3505 int ext4_calculate_overhead(struct super_block *sb) 3506 { 3507 struct ext4_sb_info *sbi = EXT4_SB(sb); 3508 struct ext4_super_block *es = sbi->s_es; 3509 struct inode *j_inode; 3510 unsigned int j_blocks, j_inum = le32_to_cpu(es->s_journal_inum); 3511 ext4_group_t i, ngroups = ext4_get_groups_count(sb); 3512 ext4_fsblk_t overhead = 0; 3513 char *buf = (char *) get_zeroed_page(GFP_NOFS); 3514 3515 if (!buf) 3516 return -ENOMEM; 3517 3518 /* 3519 * Compute the overhead (FS structures). This is constant 3520 * for a given filesystem unless the number of block groups 3521 * changes so we cache the previous value until it does. 3522 */ 3523 3524 /* 3525 * All of the blocks before first_data_block are overhead 3526 */ 3527 overhead = EXT4_B2C(sbi, le32_to_cpu(es->s_first_data_block)); 3528 3529 /* 3530 * Add the overhead found in each block group 3531 */ 3532 for (i = 0; i < ngroups; i++) { 3533 int blks; 3534 3535 blks = count_overhead(sb, i, buf); 3536 overhead += blks; 3537 if (blks) 3538 memset(buf, 0, PAGE_SIZE); 3539 cond_resched(); 3540 } 3541 3542 /* 3543 * Add the internal journal blocks whether the journal has been 3544 * loaded or not 3545 */ 3546 if (sbi->s_journal && !sbi->journal_bdev) 3547 overhead += EXT4_NUM_B2C(sbi, sbi->s_journal->j_maxlen); 3548 else if (ext4_has_feature_journal(sb) && !sbi->s_journal) { 3549 j_inode = ext4_get_journal_inode(sb, j_inum); 3550 if (j_inode) { 3551 j_blocks = j_inode->i_size >> sb->s_blocksize_bits; 3552 overhead += EXT4_NUM_B2C(sbi, j_blocks); 3553 iput(j_inode); 3554 } else { 3555 ext4_msg(sb, KERN_ERR, "can't get journal size"); 3556 } 3557 } 3558 sbi->s_overhead = overhead; 3559 smp_wmb(); 3560 free_page((unsigned long) buf); 3561 return 0; 3562 } 3563 3564 static void ext4_set_resv_clusters(struct super_block *sb) 3565 { 3566 ext4_fsblk_t resv_clusters; 3567 struct ext4_sb_info *sbi = EXT4_SB(sb); 3568 3569 /* 3570 * There's no need to reserve anything when we aren't using extents. 3571 * The space estimates are exact, there are no unwritten extents, 3572 * hole punching doesn't need new metadata... This is needed especially 3573 * to keep ext2/3 backward compatibility. 3574 */ 3575 if (!ext4_has_feature_extents(sb)) 3576 return; 3577 /* 3578 * By default we reserve 2% or 4096 clusters, whichever is smaller. 3579 * This should cover the situations where we can not afford to run 3580 * out of space like for example punch hole, or converting 3581 * unwritten extents in delalloc path. In most cases such 3582 * allocation would require 1, or 2 blocks, higher numbers are 3583 * very rare. 3584 */ 3585 resv_clusters = (ext4_blocks_count(sbi->s_es) >> 3586 sbi->s_cluster_bits); 3587 3588 do_div(resv_clusters, 50); 3589 resv_clusters = min_t(ext4_fsblk_t, resv_clusters, 4096); 3590 3591 atomic64_set(&sbi->s_resv_clusters, resv_clusters); 3592 } 3593 3594 static int ext4_fill_super(struct super_block *sb, void *data, int silent) 3595 { 3596 struct dax_device *dax_dev = fs_dax_get_by_bdev(sb->s_bdev); 3597 char *orig_data = kstrdup(data, GFP_KERNEL); 3598 struct buffer_head *bh; 3599 struct ext4_super_block *es = NULL; 3600 struct ext4_sb_info *sbi = kzalloc(sizeof(*sbi), GFP_KERNEL); 3601 ext4_fsblk_t block; 3602 ext4_fsblk_t sb_block = get_sb_block(&data); 3603 ext4_fsblk_t logical_sb_block; 3604 unsigned long offset = 0; 3605 unsigned long journal_devnum = 0; 3606 unsigned long def_mount_opts; 3607 struct inode *root; 3608 const char *descr; 3609 int ret = -ENOMEM; 3610 int blocksize, clustersize; 3611 unsigned int db_count; 3612 unsigned int i; 3613 int needs_recovery, has_huge_files, has_bigalloc; 3614 __u64 blocks_count; 3615 int err = 0; 3616 unsigned int journal_ioprio = DEFAULT_JOURNAL_IOPRIO; 3617 ext4_group_t first_not_zeroed; 3618 3619 if ((data && !orig_data) || !sbi) 3620 goto out_free_base; 3621 3622 sbi->s_daxdev = dax_dev; 3623 sbi->s_blockgroup_lock = 3624 kzalloc(sizeof(struct blockgroup_lock), GFP_KERNEL); 3625 if (!sbi->s_blockgroup_lock) 3626 goto out_free_base; 3627 3628 sb->s_fs_info = sbi; 3629 sbi->s_sb = sb; 3630 sbi->s_inode_readahead_blks = EXT4_DEF_INODE_READAHEAD_BLKS; 3631 sbi->s_sb_block = sb_block; 3632 if (sb->s_bdev->bd_part) 3633 sbi->s_sectors_written_start = 3634 part_stat_read(sb->s_bdev->bd_part, sectors[STAT_WRITE]); 3635 3636 /* Cleanup superblock name */ 3637 strreplace(sb->s_id, '/', '!'); 3638 3639 /* -EINVAL is default */ 3640 ret = -EINVAL; 3641 blocksize = sb_min_blocksize(sb, EXT4_MIN_BLOCK_SIZE); 3642 if (!blocksize) { 3643 ext4_msg(sb, KERN_ERR, "unable to set blocksize"); 3644 goto out_fail; 3645 } 3646 3647 /* 3648 * The ext4 superblock will not be buffer aligned for other than 1kB 3649 * block sizes. We need to calculate the offset from buffer start. 3650 */ 3651 if (blocksize != EXT4_MIN_BLOCK_SIZE) { 3652 logical_sb_block = sb_block * EXT4_MIN_BLOCK_SIZE; 3653 offset = do_div(logical_sb_block, blocksize); 3654 } else { 3655 logical_sb_block = sb_block; 3656 } 3657 3658 if (!(bh = sb_bread_unmovable(sb, logical_sb_block))) { 3659 ext4_msg(sb, KERN_ERR, "unable to read superblock"); 3660 goto out_fail; 3661 } 3662 /* 3663 * Note: s_es must be initialized as soon as possible because 3664 * some ext4 macro-instructions depend on its value 3665 */ 3666 es = (struct ext4_super_block *) (bh->b_data + offset); 3667 sbi->s_es = es; 3668 sb->s_magic = le16_to_cpu(es->s_magic); 3669 if (sb->s_magic != EXT4_SUPER_MAGIC) 3670 goto cantfind_ext4; 3671 sbi->s_kbytes_written = le64_to_cpu(es->s_kbytes_written); 3672 3673 /* Warn if metadata_csum and gdt_csum are both set. */ 3674 if (ext4_has_feature_metadata_csum(sb) && 3675 ext4_has_feature_gdt_csum(sb)) 3676 ext4_warning(sb, "metadata_csum and uninit_bg are " 3677 "redundant flags; please run fsck."); 3678 3679 /* Check for a known checksum algorithm */ 3680 if (!ext4_verify_csum_type(sb, es)) { 3681 ext4_msg(sb, KERN_ERR, "VFS: Found ext4 filesystem with " 3682 "unknown checksum algorithm."); 3683 silent = 1; 3684 goto cantfind_ext4; 3685 } 3686 3687 /* Load the checksum driver */ 3688 sbi->s_chksum_driver = crypto_alloc_shash("crc32c", 0, 0); 3689 if (IS_ERR(sbi->s_chksum_driver)) { 3690 ext4_msg(sb, KERN_ERR, "Cannot load crc32c driver."); 3691 ret = PTR_ERR(sbi->s_chksum_driver); 3692 sbi->s_chksum_driver = NULL; 3693 goto failed_mount; 3694 } 3695 3696 /* Check superblock checksum */ 3697 if (!ext4_superblock_csum_verify(sb, es)) { 3698 ext4_msg(sb, KERN_ERR, "VFS: Found ext4 filesystem with " 3699 "invalid superblock checksum. Run e2fsck?"); 3700 silent = 1; 3701 ret = -EFSBADCRC; 3702 goto cantfind_ext4; 3703 } 3704 3705 /* Precompute checksum seed for all metadata */ 3706 if (ext4_has_feature_csum_seed(sb)) 3707 sbi->s_csum_seed = le32_to_cpu(es->s_checksum_seed); 3708 else if (ext4_has_metadata_csum(sb) || ext4_has_feature_ea_inode(sb)) 3709 sbi->s_csum_seed = ext4_chksum(sbi, ~0, es->s_uuid, 3710 sizeof(es->s_uuid)); 3711 3712 /* Set defaults before we parse the mount options */ 3713 def_mount_opts = le32_to_cpu(es->s_default_mount_opts); 3714 set_opt(sb, INIT_INODE_TABLE); 3715 if (def_mount_opts & EXT4_DEFM_DEBUG) 3716 set_opt(sb, DEBUG); 3717 if (def_mount_opts & EXT4_DEFM_BSDGROUPS) 3718 set_opt(sb, GRPID); 3719 if (def_mount_opts & EXT4_DEFM_UID16) 3720 set_opt(sb, NO_UID32); 3721 /* xattr user namespace & acls are now defaulted on */ 3722 set_opt(sb, XATTR_USER); 3723 #ifdef CONFIG_EXT4_FS_POSIX_ACL 3724 set_opt(sb, POSIX_ACL); 3725 #endif 3726 /* don't forget to enable journal_csum when metadata_csum is enabled. */ 3727 if (ext4_has_metadata_csum(sb)) 3728 set_opt(sb, JOURNAL_CHECKSUM); 3729 3730 if ((def_mount_opts & EXT4_DEFM_JMODE) == EXT4_DEFM_JMODE_DATA) 3731 set_opt(sb, JOURNAL_DATA); 3732 else if ((def_mount_opts & EXT4_DEFM_JMODE) == EXT4_DEFM_JMODE_ORDERED) 3733 set_opt(sb, ORDERED_DATA); 3734 else if ((def_mount_opts & EXT4_DEFM_JMODE) == EXT4_DEFM_JMODE_WBACK) 3735 set_opt(sb, WRITEBACK_DATA); 3736 3737 if (le16_to_cpu(sbi->s_es->s_errors) == EXT4_ERRORS_PANIC) 3738 set_opt(sb, ERRORS_PANIC); 3739 else if (le16_to_cpu(sbi->s_es->s_errors) == EXT4_ERRORS_CONTINUE) 3740 set_opt(sb, ERRORS_CONT); 3741 else 3742 set_opt(sb, ERRORS_RO); 3743 /* block_validity enabled by default; disable with noblock_validity */ 3744 set_opt(sb, BLOCK_VALIDITY); 3745 if (def_mount_opts & EXT4_DEFM_DISCARD) 3746 set_opt(sb, DISCARD); 3747 3748 sbi->s_resuid = make_kuid(&init_user_ns, le16_to_cpu(es->s_def_resuid)); 3749 sbi->s_resgid = make_kgid(&init_user_ns, le16_to_cpu(es->s_def_resgid)); 3750 sbi->s_commit_interval = JBD2_DEFAULT_MAX_COMMIT_AGE * HZ; 3751 sbi->s_min_batch_time = EXT4_DEF_MIN_BATCH_TIME; 3752 sbi->s_max_batch_time = EXT4_DEF_MAX_BATCH_TIME; 3753 3754 if ((def_mount_opts & EXT4_DEFM_NOBARRIER) == 0) 3755 set_opt(sb, BARRIER); 3756 3757 /* 3758 * enable delayed allocation by default 3759 * Use -o nodelalloc to turn it off 3760 */ 3761 if (!IS_EXT3_SB(sb) && !IS_EXT2_SB(sb) && 3762 ((def_mount_opts & EXT4_DEFM_NODELALLOC) == 0)) 3763 set_opt(sb, DELALLOC); 3764 3765 /* 3766 * set default s_li_wait_mult for lazyinit, for the case there is 3767 * no mount option specified. 3768 */ 3769 sbi->s_li_wait_mult = EXT4_DEF_LI_WAIT_MULT; 3770 3771 if (le32_to_cpu(es->s_rev_level) == EXT4_GOOD_OLD_REV) { 3772 sbi->s_inode_size = EXT4_GOOD_OLD_INODE_SIZE; 3773 sbi->s_first_ino = EXT4_GOOD_OLD_FIRST_INO; 3774 } else { 3775 sbi->s_inode_size = le16_to_cpu(es->s_inode_size); 3776 sbi->s_first_ino = le32_to_cpu(es->s_first_ino); 3777 if (sbi->s_first_ino < EXT4_GOOD_OLD_FIRST_INO) { 3778 ext4_msg(sb, KERN_ERR, "invalid first ino: %u", 3779 sbi->s_first_ino); 3780 goto failed_mount; 3781 } 3782 if ((sbi->s_inode_size < EXT4_GOOD_OLD_INODE_SIZE) || 3783 (!is_power_of_2(sbi->s_inode_size)) || 3784 (sbi->s_inode_size > blocksize)) { 3785 ext4_msg(sb, KERN_ERR, 3786 "unsupported inode size: %d", 3787 sbi->s_inode_size); 3788 goto failed_mount; 3789 } 3790 /* 3791 * i_atime_extra is the last extra field available for 3792 * [acm]times in struct ext4_inode. Checking for that 3793 * field should suffice to ensure we have extra space 3794 * for all three. 3795 */ 3796 if (sbi->s_inode_size >= offsetof(struct ext4_inode, i_atime_extra) + 3797 sizeof(((struct ext4_inode *)0)->i_atime_extra)) { 3798 sb->s_time_gran = 1; 3799 sb->s_time_max = EXT4_EXTRA_TIMESTAMP_MAX; 3800 } else { 3801 sb->s_time_gran = NSEC_PER_SEC; 3802 sb->s_time_max = EXT4_NON_EXTRA_TIMESTAMP_MAX; 3803 } 3804 sb->s_time_min = EXT4_TIMESTAMP_MIN; 3805 } 3806 if (sbi->s_inode_size > EXT4_GOOD_OLD_INODE_SIZE) { 3807 sbi->s_want_extra_isize = sizeof(struct ext4_inode) - 3808 EXT4_GOOD_OLD_INODE_SIZE; 3809 if (ext4_has_feature_extra_isize(sb)) { 3810 unsigned v, max = (sbi->s_inode_size - 3811 EXT4_GOOD_OLD_INODE_SIZE); 3812 3813 v = le16_to_cpu(es->s_want_extra_isize); 3814 if (v > max) { 3815 ext4_msg(sb, KERN_ERR, 3816 "bad s_want_extra_isize: %d", v); 3817 goto failed_mount; 3818 } 3819 if (sbi->s_want_extra_isize < v) 3820 sbi->s_want_extra_isize = v; 3821 3822 v = le16_to_cpu(es->s_min_extra_isize); 3823 if (v > max) { 3824 ext4_msg(sb, KERN_ERR, 3825 "bad s_min_extra_isize: %d", v); 3826 goto failed_mount; 3827 } 3828 if (sbi->s_want_extra_isize < v) 3829 sbi->s_want_extra_isize = v; 3830 } 3831 } 3832 3833 if (sbi->s_es->s_mount_opts[0]) { 3834 char *s_mount_opts = kstrndup(sbi->s_es->s_mount_opts, 3835 sizeof(sbi->s_es->s_mount_opts), 3836 GFP_KERNEL); 3837 if (!s_mount_opts) 3838 goto failed_mount; 3839 if (!parse_options(s_mount_opts, sb, &journal_devnum, 3840 &journal_ioprio, 0)) { 3841 ext4_msg(sb, KERN_WARNING, 3842 "failed to parse options in superblock: %s", 3843 s_mount_opts); 3844 } 3845 kfree(s_mount_opts); 3846 } 3847 sbi->s_def_mount_opt = sbi->s_mount_opt; 3848 if (!parse_options((char *) data, sb, &journal_devnum, 3849 &journal_ioprio, 0)) 3850 goto failed_mount; 3851 3852 #ifdef CONFIG_UNICODE 3853 if (ext4_has_feature_casefold(sb) && !sbi->s_encoding) { 3854 const struct ext4_sb_encodings *encoding_info; 3855 struct unicode_map *encoding; 3856 __u16 encoding_flags; 3857 3858 if (ext4_has_feature_encrypt(sb)) { 3859 ext4_msg(sb, KERN_ERR, 3860 "Can't mount with encoding and encryption"); 3861 goto failed_mount; 3862 } 3863 3864 if (ext4_sb_read_encoding(es, &encoding_info, 3865 &encoding_flags)) { 3866 ext4_msg(sb, KERN_ERR, 3867 "Encoding requested by superblock is unknown"); 3868 goto failed_mount; 3869 } 3870 3871 encoding = utf8_load(encoding_info->version); 3872 if (IS_ERR(encoding)) { 3873 ext4_msg(sb, KERN_ERR, 3874 "can't mount with superblock charset: %s-%s " 3875 "not supported by the kernel. flags: 0x%x.", 3876 encoding_info->name, encoding_info->version, 3877 encoding_flags); 3878 goto failed_mount; 3879 } 3880 ext4_msg(sb, KERN_INFO,"Using encoding defined by superblock: " 3881 "%s-%s with flags 0x%hx", encoding_info->name, 3882 encoding_info->version?:"\b", encoding_flags); 3883 3884 sbi->s_encoding = encoding; 3885 sbi->s_encoding_flags = encoding_flags; 3886 } 3887 #endif 3888 3889 if (test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_JOURNAL_DATA) { 3890 printk_once(KERN_WARNING "EXT4-fs: Warning: mounting " 3891 "with data=journal disables delayed " 3892 "allocation and O_DIRECT support!\n"); 3893 if (test_opt2(sb, EXPLICIT_DELALLOC)) { 3894 ext4_msg(sb, KERN_ERR, "can't mount with " 3895 "both data=journal and delalloc"); 3896 goto failed_mount; 3897 } 3898 if (test_opt(sb, DIOREAD_NOLOCK)) { 3899 ext4_msg(sb, KERN_ERR, "can't mount with " 3900 "both data=journal and dioread_nolock"); 3901 goto failed_mount; 3902 } 3903 if (test_opt(sb, DAX)) { 3904 ext4_msg(sb, KERN_ERR, "can't mount with " 3905 "both data=journal and dax"); 3906 goto failed_mount; 3907 } 3908 if (ext4_has_feature_encrypt(sb)) { 3909 ext4_msg(sb, KERN_WARNING, 3910 "encrypted files will use data=ordered " 3911 "instead of data journaling mode"); 3912 } 3913 if (test_opt(sb, DELALLOC)) 3914 clear_opt(sb, DELALLOC); 3915 } else { 3916 sb->s_iflags |= SB_I_CGROUPWB; 3917 } 3918 3919 sb->s_flags = (sb->s_flags & ~SB_POSIXACL) | 3920 (test_opt(sb, POSIX_ACL) ? SB_POSIXACL : 0); 3921 3922 if (le32_to_cpu(es->s_rev_level) == EXT4_GOOD_OLD_REV && 3923 (ext4_has_compat_features(sb) || 3924 ext4_has_ro_compat_features(sb) || 3925 ext4_has_incompat_features(sb))) 3926 ext4_msg(sb, KERN_WARNING, 3927 "feature flags set on rev 0 fs, " 3928 "running e2fsck is recommended"); 3929 3930 if (es->s_creator_os == cpu_to_le32(EXT4_OS_HURD)) { 3931 set_opt2(sb, HURD_COMPAT); 3932 if (ext4_has_feature_64bit(sb)) { 3933 ext4_msg(sb, KERN_ERR, 3934 "The Hurd can't support 64-bit file systems"); 3935 goto failed_mount; 3936 } 3937 3938 /* 3939 * ea_inode feature uses l_i_version field which is not 3940 * available in HURD_COMPAT mode. 3941 */ 3942 if (ext4_has_feature_ea_inode(sb)) { 3943 ext4_msg(sb, KERN_ERR, 3944 "ea_inode feature is not supported for Hurd"); 3945 goto failed_mount; 3946 } 3947 } 3948 3949 if (IS_EXT2_SB(sb)) { 3950 if (ext2_feature_set_ok(sb)) 3951 ext4_msg(sb, KERN_INFO, "mounting ext2 file system " 3952 "using the ext4 subsystem"); 3953 else { 3954 /* 3955 * If we're probing be silent, if this looks like 3956 * it's actually an ext[34] filesystem. 3957 */ 3958 if (silent && ext4_feature_set_ok(sb, sb_rdonly(sb))) 3959 goto failed_mount; 3960 ext4_msg(sb, KERN_ERR, "couldn't mount as ext2 due " 3961 "to feature incompatibilities"); 3962 goto failed_mount; 3963 } 3964 } 3965 3966 if (IS_EXT3_SB(sb)) { 3967 if (ext3_feature_set_ok(sb)) 3968 ext4_msg(sb, KERN_INFO, "mounting ext3 file system " 3969 "using the ext4 subsystem"); 3970 else { 3971 /* 3972 * If we're probing be silent, if this looks like 3973 * it's actually an ext4 filesystem. 3974 */ 3975 if (silent && ext4_feature_set_ok(sb, sb_rdonly(sb))) 3976 goto failed_mount; 3977 ext4_msg(sb, KERN_ERR, "couldn't mount as ext3 due " 3978 "to feature incompatibilities"); 3979 goto failed_mount; 3980 } 3981 } 3982 3983 /* 3984 * Check feature flags regardless of the revision level, since we 3985 * previously didn't change the revision level when setting the flags, 3986 * so there is a chance incompat flags are set on a rev 0 filesystem. 3987 */ 3988 if (!ext4_feature_set_ok(sb, (sb_rdonly(sb)))) 3989 goto failed_mount; 3990 3991 blocksize = BLOCK_SIZE << le32_to_cpu(es->s_log_block_size); 3992 if (blocksize < EXT4_MIN_BLOCK_SIZE || 3993 blocksize > EXT4_MAX_BLOCK_SIZE) { 3994 ext4_msg(sb, KERN_ERR, 3995 "Unsupported filesystem blocksize %d (%d log_block_size)", 3996 blocksize, le32_to_cpu(es->s_log_block_size)); 3997 goto failed_mount; 3998 } 3999 if (le32_to_cpu(es->s_log_block_size) > 4000 (EXT4_MAX_BLOCK_LOG_SIZE - EXT4_MIN_BLOCK_LOG_SIZE)) { 4001 ext4_msg(sb, KERN_ERR, 4002 "Invalid log block size: %u", 4003 le32_to_cpu(es->s_log_block_size)); 4004 goto failed_mount; 4005 } 4006 if (le32_to_cpu(es->s_log_cluster_size) > 4007 (EXT4_MAX_CLUSTER_LOG_SIZE - EXT4_MIN_BLOCK_LOG_SIZE)) { 4008 ext4_msg(sb, KERN_ERR, 4009 "Invalid log cluster size: %u", 4010 le32_to_cpu(es->s_log_cluster_size)); 4011 goto failed_mount; 4012 } 4013 4014 if (le16_to_cpu(sbi->s_es->s_reserved_gdt_blocks) > (blocksize / 4)) { 4015 ext4_msg(sb, KERN_ERR, 4016 "Number of reserved GDT blocks insanely large: %d", 4017 le16_to_cpu(sbi->s_es->s_reserved_gdt_blocks)); 4018 goto failed_mount; 4019 } 4020 4021 if (sbi->s_mount_opt & EXT4_MOUNT_DAX) { 4022 if (ext4_has_feature_inline_data(sb)) { 4023 ext4_msg(sb, KERN_ERR, "Cannot use DAX on a filesystem" 4024 " that may contain inline data"); 4025 goto failed_mount; 4026 } 4027 if (!bdev_dax_supported(sb->s_bdev, blocksize)) { 4028 ext4_msg(sb, KERN_ERR, 4029 "DAX unsupported by block device."); 4030 goto failed_mount; 4031 } 4032 } 4033 4034 if (ext4_has_feature_encrypt(sb) && es->s_encryption_level) { 4035 ext4_msg(sb, KERN_ERR, "Unsupported encryption level %d", 4036 es->s_encryption_level); 4037 goto failed_mount; 4038 } 4039 4040 if (sb->s_blocksize != blocksize) { 4041 /* Validate the filesystem blocksize */ 4042 if (!sb_set_blocksize(sb, blocksize)) { 4043 ext4_msg(sb, KERN_ERR, "bad block size %d", 4044 blocksize); 4045 goto failed_mount; 4046 } 4047 4048 brelse(bh); 4049 logical_sb_block = sb_block * EXT4_MIN_BLOCK_SIZE; 4050 offset = do_div(logical_sb_block, blocksize); 4051 bh = sb_bread_unmovable(sb, logical_sb_block); 4052 if (!bh) { 4053 ext4_msg(sb, KERN_ERR, 4054 "Can't read superblock on 2nd try"); 4055 goto failed_mount; 4056 } 4057 es = (struct ext4_super_block *)(bh->b_data + offset); 4058 sbi->s_es = es; 4059 if (es->s_magic != cpu_to_le16(EXT4_SUPER_MAGIC)) { 4060 ext4_msg(sb, KERN_ERR, 4061 "Magic mismatch, very weird!"); 4062 goto failed_mount; 4063 } 4064 } 4065 4066 has_huge_files = ext4_has_feature_huge_file(sb); 4067 sbi->s_bitmap_maxbytes = ext4_max_bitmap_size(sb->s_blocksize_bits, 4068 has_huge_files); 4069 sb->s_maxbytes = ext4_max_size(sb->s_blocksize_bits, has_huge_files); 4070 4071 sbi->s_desc_size = le16_to_cpu(es->s_desc_size); 4072 if (ext4_has_feature_64bit(sb)) { 4073 if (sbi->s_desc_size < EXT4_MIN_DESC_SIZE_64BIT || 4074 sbi->s_desc_size > EXT4_MAX_DESC_SIZE || 4075 !is_power_of_2(sbi->s_desc_size)) { 4076 ext4_msg(sb, KERN_ERR, 4077 "unsupported descriptor size %lu", 4078 sbi->s_desc_size); 4079 goto failed_mount; 4080 } 4081 } else 4082 sbi->s_desc_size = EXT4_MIN_DESC_SIZE; 4083 4084 sbi->s_blocks_per_group = le32_to_cpu(es->s_blocks_per_group); 4085 sbi->s_inodes_per_group = le32_to_cpu(es->s_inodes_per_group); 4086 4087 sbi->s_inodes_per_block = blocksize / EXT4_INODE_SIZE(sb); 4088 if (sbi->s_inodes_per_block == 0) 4089 goto cantfind_ext4; 4090 if (sbi->s_inodes_per_group < sbi->s_inodes_per_block || 4091 sbi->s_inodes_per_group > blocksize * 8) { 4092 ext4_msg(sb, KERN_ERR, "invalid inodes per group: %lu\n", 4093 sbi->s_blocks_per_group); 4094 goto failed_mount; 4095 } 4096 sbi->s_itb_per_group = sbi->s_inodes_per_group / 4097 sbi->s_inodes_per_block; 4098 sbi->s_desc_per_block = blocksize / EXT4_DESC_SIZE(sb); 4099 sbi->s_sbh = bh; 4100 sbi->s_mount_state = le16_to_cpu(es->s_state); 4101 sbi->s_addr_per_block_bits = ilog2(EXT4_ADDR_PER_BLOCK(sb)); 4102 sbi->s_desc_per_block_bits = ilog2(EXT4_DESC_PER_BLOCK(sb)); 4103 4104 for (i = 0; i < 4; i++) 4105 sbi->s_hash_seed[i] = le32_to_cpu(es->s_hash_seed[i]); 4106 sbi->s_def_hash_version = es->s_def_hash_version; 4107 if (ext4_has_feature_dir_index(sb)) { 4108 i = le32_to_cpu(es->s_flags); 4109 if (i & EXT2_FLAGS_UNSIGNED_HASH) 4110 sbi->s_hash_unsigned = 3; 4111 else if ((i & EXT2_FLAGS_SIGNED_HASH) == 0) { 4112 #ifdef __CHAR_UNSIGNED__ 4113 if (!sb_rdonly(sb)) 4114 es->s_flags |= 4115 cpu_to_le32(EXT2_FLAGS_UNSIGNED_HASH); 4116 sbi->s_hash_unsigned = 3; 4117 #else 4118 if (!sb_rdonly(sb)) 4119 es->s_flags |= 4120 cpu_to_le32(EXT2_FLAGS_SIGNED_HASH); 4121 #endif 4122 } 4123 } 4124 4125 /* Handle clustersize */ 4126 clustersize = BLOCK_SIZE << le32_to_cpu(es->s_log_cluster_size); 4127 has_bigalloc = ext4_has_feature_bigalloc(sb); 4128 if (has_bigalloc) { 4129 if (clustersize < blocksize) { 4130 ext4_msg(sb, KERN_ERR, 4131 "cluster size (%d) smaller than " 4132 "block size (%d)", clustersize, blocksize); 4133 goto failed_mount; 4134 } 4135 sbi->s_cluster_bits = le32_to_cpu(es->s_log_cluster_size) - 4136 le32_to_cpu(es->s_log_block_size); 4137 sbi->s_clusters_per_group = 4138 le32_to_cpu(es->s_clusters_per_group); 4139 if (sbi->s_clusters_per_group > blocksize * 8) { 4140 ext4_msg(sb, KERN_ERR, 4141 "#clusters per group too big: %lu", 4142 sbi->s_clusters_per_group); 4143 goto failed_mount; 4144 } 4145 if (sbi->s_blocks_per_group != 4146 (sbi->s_clusters_per_group * (clustersize / blocksize))) { 4147 ext4_msg(sb, KERN_ERR, "blocks per group (%lu) and " 4148 "clusters per group (%lu) inconsistent", 4149 sbi->s_blocks_per_group, 4150 sbi->s_clusters_per_group); 4151 goto failed_mount; 4152 } 4153 } else { 4154 if (clustersize != blocksize) { 4155 ext4_msg(sb, KERN_ERR, 4156 "fragment/cluster size (%d) != " 4157 "block size (%d)", clustersize, blocksize); 4158 goto failed_mount; 4159 } 4160 if (sbi->s_blocks_per_group > blocksize * 8) { 4161 ext4_msg(sb, KERN_ERR, 4162 "#blocks per group too big: %lu", 4163 sbi->s_blocks_per_group); 4164 goto failed_mount; 4165 } 4166 sbi->s_clusters_per_group = sbi->s_blocks_per_group; 4167 sbi->s_cluster_bits = 0; 4168 } 4169 sbi->s_cluster_ratio = clustersize / blocksize; 4170 4171 /* Do we have standard group size of clustersize * 8 blocks ? */ 4172 if (sbi->s_blocks_per_group == clustersize << 3) 4173 set_opt2(sb, STD_GROUP_SIZE); 4174 4175 /* 4176 * Test whether we have more sectors than will fit in sector_t, 4177 * and whether the max offset is addressable by the page cache. 4178 */ 4179 err = generic_check_addressable(sb->s_blocksize_bits, 4180 ext4_blocks_count(es)); 4181 if (err) { 4182 ext4_msg(sb, KERN_ERR, "filesystem" 4183 " too large to mount safely on this system"); 4184 goto failed_mount; 4185 } 4186 4187 if (EXT4_BLOCKS_PER_GROUP(sb) == 0) 4188 goto cantfind_ext4; 4189 4190 /* check blocks count against device size */ 4191 blocks_count = sb->s_bdev->bd_inode->i_size >> sb->s_blocksize_bits; 4192 if (blocks_count && ext4_blocks_count(es) > blocks_count) { 4193 ext4_msg(sb, KERN_WARNING, "bad geometry: block count %llu " 4194 "exceeds size of device (%llu blocks)", 4195 ext4_blocks_count(es), blocks_count); 4196 goto failed_mount; 4197 } 4198 4199 /* 4200 * It makes no sense for the first data block to be beyond the end 4201 * of the filesystem. 4202 */ 4203 if (le32_to_cpu(es->s_first_data_block) >= ext4_blocks_count(es)) { 4204 ext4_msg(sb, KERN_WARNING, "bad geometry: first data " 4205 "block %u is beyond end of filesystem (%llu)", 4206 le32_to_cpu(es->s_first_data_block), 4207 ext4_blocks_count(es)); 4208 goto failed_mount; 4209 } 4210 if ((es->s_first_data_block == 0) && (es->s_log_block_size == 0) && 4211 (sbi->s_cluster_ratio == 1)) { 4212 ext4_msg(sb, KERN_WARNING, "bad geometry: first data " 4213 "block is 0 with a 1k block and cluster size"); 4214 goto failed_mount; 4215 } 4216 4217 blocks_count = (ext4_blocks_count(es) - 4218 le32_to_cpu(es->s_first_data_block) + 4219 EXT4_BLOCKS_PER_GROUP(sb) - 1); 4220 do_div(blocks_count, EXT4_BLOCKS_PER_GROUP(sb)); 4221 if (blocks_count > ((uint64_t)1<<32) - EXT4_DESC_PER_BLOCK(sb)) { 4222 ext4_msg(sb, KERN_WARNING, "groups count too large: %u " 4223 "(block count %llu, first data block %u, " 4224 "blocks per group %lu)", sbi->s_groups_count, 4225 ext4_blocks_count(es), 4226 le32_to_cpu(es->s_first_data_block), 4227 EXT4_BLOCKS_PER_GROUP(sb)); 4228 goto failed_mount; 4229 } 4230 sbi->s_groups_count = blocks_count; 4231 sbi->s_blockfile_groups = min_t(ext4_group_t, sbi->s_groups_count, 4232 (EXT4_MAX_BLOCK_FILE_PHYS / EXT4_BLOCKS_PER_GROUP(sb))); 4233 if (((u64)sbi->s_groups_count * sbi->s_inodes_per_group) != 4234 le32_to_cpu(es->s_inodes_count)) { 4235 ext4_msg(sb, KERN_ERR, "inodes count not valid: %u vs %llu", 4236 le32_to_cpu(es->s_inodes_count), 4237 ((u64)sbi->s_groups_count * sbi->s_inodes_per_group)); 4238 ret = -EINVAL; 4239 goto failed_mount; 4240 } 4241 db_count = (sbi->s_groups_count + EXT4_DESC_PER_BLOCK(sb) - 1) / 4242 EXT4_DESC_PER_BLOCK(sb); 4243 if (ext4_has_feature_meta_bg(sb)) { 4244 if (le32_to_cpu(es->s_first_meta_bg) > db_count) { 4245 ext4_msg(sb, KERN_WARNING, 4246 "first meta block group too large: %u " 4247 "(group descriptor block count %u)", 4248 le32_to_cpu(es->s_first_meta_bg), db_count); 4249 goto failed_mount; 4250 } 4251 } 4252 sbi->s_group_desc = kvmalloc_array(db_count, 4253 sizeof(struct buffer_head *), 4254 GFP_KERNEL); 4255 if (sbi->s_group_desc == NULL) { 4256 ext4_msg(sb, KERN_ERR, "not enough memory"); 4257 ret = -ENOMEM; 4258 goto failed_mount; 4259 } 4260 4261 bgl_lock_init(sbi->s_blockgroup_lock); 4262 4263 /* Pre-read the descriptors into the buffer cache */ 4264 for (i = 0; i < db_count; i++) { 4265 block = descriptor_loc(sb, logical_sb_block, i); 4266 sb_breadahead(sb, block); 4267 } 4268 4269 for (i = 0; i < db_count; i++) { 4270 block = descriptor_loc(sb, logical_sb_block, i); 4271 sbi->s_group_desc[i] = sb_bread_unmovable(sb, block); 4272 if (!sbi->s_group_desc[i]) { 4273 ext4_msg(sb, KERN_ERR, 4274 "can't read group descriptor %d", i); 4275 db_count = i; 4276 goto failed_mount2; 4277 } 4278 } 4279 sbi->s_gdb_count = db_count; 4280 if (!ext4_check_descriptors(sb, logical_sb_block, &first_not_zeroed)) { 4281 ext4_msg(sb, KERN_ERR, "group descriptors corrupted!"); 4282 ret = -EFSCORRUPTED; 4283 goto failed_mount2; 4284 } 4285 4286 timer_setup(&sbi->s_err_report, print_daily_error_info, 0); 4287 4288 /* Register extent status tree shrinker */ 4289 if (ext4_es_register_shrinker(sbi)) 4290 goto failed_mount3; 4291 4292 sbi->s_stripe = ext4_get_stripe_size(sbi); 4293 sbi->s_extent_max_zeroout_kb = 32; 4294 4295 /* 4296 * set up enough so that it can read an inode 4297 */ 4298 sb->s_op = &ext4_sops; 4299 sb->s_export_op = &ext4_export_ops; 4300 sb->s_xattr = ext4_xattr_handlers; 4301 #ifdef CONFIG_FS_ENCRYPTION 4302 sb->s_cop = &ext4_cryptops; 4303 #endif 4304 #ifdef CONFIG_FS_VERITY 4305 sb->s_vop = &ext4_verityops; 4306 #endif 4307 #ifdef CONFIG_QUOTA 4308 sb->dq_op = &ext4_quota_operations; 4309 if (ext4_has_feature_quota(sb)) 4310 sb->s_qcop = &dquot_quotactl_sysfile_ops; 4311 else 4312 sb->s_qcop = &ext4_qctl_operations; 4313 sb->s_quota_types = QTYPE_MASK_USR | QTYPE_MASK_GRP | QTYPE_MASK_PRJ; 4314 #endif 4315 memcpy(&sb->s_uuid, es->s_uuid, sizeof(es->s_uuid)); 4316 4317 INIT_LIST_HEAD(&sbi->s_orphan); /* unlinked but open files */ 4318 mutex_init(&sbi->s_orphan_lock); 4319 4320 sb->s_root = NULL; 4321 4322 needs_recovery = (es->s_last_orphan != 0 || 4323 ext4_has_feature_journal_needs_recovery(sb)); 4324 4325 if (ext4_has_feature_mmp(sb) && !sb_rdonly(sb)) 4326 if (ext4_multi_mount_protect(sb, le64_to_cpu(es->s_mmp_block))) 4327 goto failed_mount3a; 4328 4329 /* 4330 * The first inode we look at is the journal inode. Don't try 4331 * root first: it may be modified in the journal! 4332 */ 4333 if (!test_opt(sb, NOLOAD) && ext4_has_feature_journal(sb)) { 4334 err = ext4_load_journal(sb, es, journal_devnum); 4335 if (err) 4336 goto failed_mount3a; 4337 } else if (test_opt(sb, NOLOAD) && !sb_rdonly(sb) && 4338 ext4_has_feature_journal_needs_recovery(sb)) { 4339 ext4_msg(sb, KERN_ERR, "required journal recovery " 4340 "suppressed and not mounted read-only"); 4341 goto failed_mount_wq; 4342 } else { 4343 /* Nojournal mode, all journal mount options are illegal */ 4344 if (test_opt2(sb, EXPLICIT_JOURNAL_CHECKSUM)) { 4345 ext4_msg(sb, KERN_ERR, "can't mount with " 4346 "journal_checksum, fs mounted w/o journal"); 4347 goto failed_mount_wq; 4348 } 4349 if (test_opt(sb, JOURNAL_ASYNC_COMMIT)) { 4350 ext4_msg(sb, KERN_ERR, "can't mount with " 4351 "journal_async_commit, fs mounted w/o journal"); 4352 goto failed_mount_wq; 4353 } 4354 if (sbi->s_commit_interval != JBD2_DEFAULT_MAX_COMMIT_AGE*HZ) { 4355 ext4_msg(sb, KERN_ERR, "can't mount with " 4356 "commit=%lu, fs mounted w/o journal", 4357 sbi->s_commit_interval / HZ); 4358 goto failed_mount_wq; 4359 } 4360 if (EXT4_MOUNT_DATA_FLAGS & 4361 (sbi->s_mount_opt ^ sbi->s_def_mount_opt)) { 4362 ext4_msg(sb, KERN_ERR, "can't mount with " 4363 "data=, fs mounted w/o journal"); 4364 goto failed_mount_wq; 4365 } 4366 sbi->s_def_mount_opt &= ~EXT4_MOUNT_JOURNAL_CHECKSUM; 4367 clear_opt(sb, JOURNAL_CHECKSUM); 4368 clear_opt(sb, DATA_FLAGS); 4369 sbi->s_journal = NULL; 4370 needs_recovery = 0; 4371 goto no_journal; 4372 } 4373 4374 if (ext4_has_feature_64bit(sb) && 4375 !jbd2_journal_set_features(EXT4_SB(sb)->s_journal, 0, 0, 4376 JBD2_FEATURE_INCOMPAT_64BIT)) { 4377 ext4_msg(sb, KERN_ERR, "Failed to set 64-bit journal feature"); 4378 goto failed_mount_wq; 4379 } 4380 4381 if (!set_journal_csum_feature_set(sb)) { 4382 ext4_msg(sb, KERN_ERR, "Failed to set journal checksum " 4383 "feature set"); 4384 goto failed_mount_wq; 4385 } 4386 4387 /* We have now updated the journal if required, so we can 4388 * validate the data journaling mode. */ 4389 switch (test_opt(sb, DATA_FLAGS)) { 4390 case 0: 4391 /* No mode set, assume a default based on the journal 4392 * capabilities: ORDERED_DATA if the journal can 4393 * cope, else JOURNAL_DATA 4394 */ 4395 if (jbd2_journal_check_available_features 4396 (sbi->s_journal, 0, 0, JBD2_FEATURE_INCOMPAT_REVOKE)) { 4397 set_opt(sb, ORDERED_DATA); 4398 sbi->s_def_mount_opt |= EXT4_MOUNT_ORDERED_DATA; 4399 } else { 4400 set_opt(sb, JOURNAL_DATA); 4401 sbi->s_def_mount_opt |= EXT4_MOUNT_JOURNAL_DATA; 4402 } 4403 break; 4404 4405 case EXT4_MOUNT_ORDERED_DATA: 4406 case EXT4_MOUNT_WRITEBACK_DATA: 4407 if (!jbd2_journal_check_available_features 4408 (sbi->s_journal, 0, 0, JBD2_FEATURE_INCOMPAT_REVOKE)) { 4409 ext4_msg(sb, KERN_ERR, "Journal does not support " 4410 "requested data journaling mode"); 4411 goto failed_mount_wq; 4412 } 4413 default: 4414 break; 4415 } 4416 4417 if (test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_ORDERED_DATA && 4418 test_opt(sb, JOURNAL_ASYNC_COMMIT)) { 4419 ext4_msg(sb, KERN_ERR, "can't mount with " 4420 "journal_async_commit in data=ordered mode"); 4421 goto failed_mount_wq; 4422 } 4423 4424 set_task_ioprio(sbi->s_journal->j_task, journal_ioprio); 4425 4426 sbi->s_journal->j_commit_callback = ext4_journal_commit_callback; 4427 4428 no_journal: 4429 if (!test_opt(sb, NO_MBCACHE)) { 4430 sbi->s_ea_block_cache = ext4_xattr_create_cache(); 4431 if (!sbi->s_ea_block_cache) { 4432 ext4_msg(sb, KERN_ERR, 4433 "Failed to create ea_block_cache"); 4434 goto failed_mount_wq; 4435 } 4436 4437 if (ext4_has_feature_ea_inode(sb)) { 4438 sbi->s_ea_inode_cache = ext4_xattr_create_cache(); 4439 if (!sbi->s_ea_inode_cache) { 4440 ext4_msg(sb, KERN_ERR, 4441 "Failed to create ea_inode_cache"); 4442 goto failed_mount_wq; 4443 } 4444 } 4445 } 4446 4447 if (ext4_has_feature_verity(sb) && blocksize != PAGE_SIZE) { 4448 ext4_msg(sb, KERN_ERR, "Unsupported blocksize for fs-verity"); 4449 goto failed_mount_wq; 4450 } 4451 4452 if (DUMMY_ENCRYPTION_ENABLED(sbi) && !sb_rdonly(sb) && 4453 !ext4_has_feature_encrypt(sb)) { 4454 ext4_set_feature_encrypt(sb); 4455 ext4_commit_super(sb, 1); 4456 } 4457 4458 /* 4459 * Get the # of file system overhead blocks from the 4460 * superblock if present. 4461 */ 4462 if (es->s_overhead_clusters) 4463 sbi->s_overhead = le32_to_cpu(es->s_overhead_clusters); 4464 else { 4465 err = ext4_calculate_overhead(sb); 4466 if (err) 4467 goto failed_mount_wq; 4468 } 4469 4470 /* 4471 * The maximum number of concurrent works can be high and 4472 * concurrency isn't really necessary. Limit it to 1. 4473 */ 4474 EXT4_SB(sb)->rsv_conversion_wq = 4475 alloc_workqueue("ext4-rsv-conversion", WQ_MEM_RECLAIM | WQ_UNBOUND, 1); 4476 if (!EXT4_SB(sb)->rsv_conversion_wq) { 4477 printk(KERN_ERR "EXT4-fs: failed to create workqueue\n"); 4478 ret = -ENOMEM; 4479 goto failed_mount4; 4480 } 4481 4482 /* 4483 * The jbd2_journal_load will have done any necessary log recovery, 4484 * so we can safely mount the rest of the filesystem now. 4485 */ 4486 4487 root = ext4_iget(sb, EXT4_ROOT_INO, EXT4_IGET_SPECIAL); 4488 if (IS_ERR(root)) { 4489 ext4_msg(sb, KERN_ERR, "get root inode failed"); 4490 ret = PTR_ERR(root); 4491 root = NULL; 4492 goto failed_mount4; 4493 } 4494 if (!S_ISDIR(root->i_mode) || !root->i_blocks || !root->i_size) { 4495 ext4_msg(sb, KERN_ERR, "corrupt root inode, run e2fsck"); 4496 iput(root); 4497 goto failed_mount4; 4498 } 4499 4500 #ifdef CONFIG_UNICODE 4501 if (sbi->s_encoding) 4502 sb->s_d_op = &ext4_dentry_ops; 4503 #endif 4504 4505 sb->s_root = d_make_root(root); 4506 if (!sb->s_root) { 4507 ext4_msg(sb, KERN_ERR, "get root dentry failed"); 4508 ret = -ENOMEM; 4509 goto failed_mount4; 4510 } 4511 4512 ret = ext4_setup_super(sb, es, sb_rdonly(sb)); 4513 if (ret == -EROFS) { 4514 sb->s_flags |= SB_RDONLY; 4515 ret = 0; 4516 } else if (ret) 4517 goto failed_mount4a; 4518 4519 ext4_set_resv_clusters(sb); 4520 4521 err = ext4_setup_system_zone(sb); 4522 if (err) { 4523 ext4_msg(sb, KERN_ERR, "failed to initialize system " 4524 "zone (%d)", err); 4525 goto failed_mount4a; 4526 } 4527 4528 ext4_ext_init(sb); 4529 err = ext4_mb_init(sb); 4530 if (err) { 4531 ext4_msg(sb, KERN_ERR, "failed to initialize mballoc (%d)", 4532 err); 4533 goto failed_mount5; 4534 } 4535 4536 block = ext4_count_free_clusters(sb); 4537 ext4_free_blocks_count_set(sbi->s_es, 4538 EXT4_C2B(sbi, block)); 4539 ext4_superblock_csum_set(sb); 4540 err = percpu_counter_init(&sbi->s_freeclusters_counter, block, 4541 GFP_KERNEL); 4542 if (!err) { 4543 unsigned long freei = ext4_count_free_inodes(sb); 4544 sbi->s_es->s_free_inodes_count = cpu_to_le32(freei); 4545 ext4_superblock_csum_set(sb); 4546 err = percpu_counter_init(&sbi->s_freeinodes_counter, freei, 4547 GFP_KERNEL); 4548 } 4549 if (!err) 4550 err = percpu_counter_init(&sbi->s_dirs_counter, 4551 ext4_count_dirs(sb), GFP_KERNEL); 4552 if (!err) 4553 err = percpu_counter_init(&sbi->s_dirtyclusters_counter, 0, 4554 GFP_KERNEL); 4555 if (!err) 4556 err = percpu_init_rwsem(&sbi->s_journal_flag_rwsem); 4557 4558 if (err) { 4559 ext4_msg(sb, KERN_ERR, "insufficient memory"); 4560 goto failed_mount6; 4561 } 4562 4563 if (ext4_has_feature_flex_bg(sb)) 4564 if (!ext4_fill_flex_info(sb)) { 4565 ext4_msg(sb, KERN_ERR, 4566 "unable to initialize " 4567 "flex_bg meta info!"); 4568 goto failed_mount6; 4569 } 4570 4571 err = ext4_register_li_request(sb, first_not_zeroed); 4572 if (err) 4573 goto failed_mount6; 4574 4575 err = ext4_register_sysfs(sb); 4576 if (err) 4577 goto failed_mount7; 4578 4579 #ifdef CONFIG_QUOTA 4580 /* Enable quota usage during mount. */ 4581 if (ext4_has_feature_quota(sb) && !sb_rdonly(sb)) { 4582 err = ext4_enable_quotas(sb); 4583 if (err) 4584 goto failed_mount8; 4585 } 4586 #endif /* CONFIG_QUOTA */ 4587 4588 EXT4_SB(sb)->s_mount_state |= EXT4_ORPHAN_FS; 4589 ext4_orphan_cleanup(sb, es); 4590 EXT4_SB(sb)->s_mount_state &= ~EXT4_ORPHAN_FS; 4591 if (needs_recovery) { 4592 ext4_msg(sb, KERN_INFO, "recovery complete"); 4593 ext4_mark_recovery_complete(sb, es); 4594 } 4595 if (EXT4_SB(sb)->s_journal) { 4596 if (test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_JOURNAL_DATA) 4597 descr = " journalled data mode"; 4598 else if (test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_ORDERED_DATA) 4599 descr = " ordered data mode"; 4600 else 4601 descr = " writeback data mode"; 4602 } else 4603 descr = "out journal"; 4604 4605 if (test_opt(sb, DISCARD)) { 4606 struct request_queue *q = bdev_get_queue(sb->s_bdev); 4607 if (!blk_queue_discard(q)) 4608 ext4_msg(sb, KERN_WARNING, 4609 "mounting with \"discard\" option, but " 4610 "the device does not support discard"); 4611 } 4612 4613 if (___ratelimit(&ext4_mount_msg_ratelimit, "EXT4-fs mount")) 4614 ext4_msg(sb, KERN_INFO, "mounted filesystem with%s. " 4615 "Opts: %.*s%s%s", descr, 4616 (int) sizeof(sbi->s_es->s_mount_opts), 4617 sbi->s_es->s_mount_opts, 4618 *sbi->s_es->s_mount_opts ? "; " : "", orig_data); 4619 4620 if (es->s_error_count) 4621 mod_timer(&sbi->s_err_report, jiffies + 300*HZ); /* 5 minutes */ 4622 4623 /* Enable message ratelimiting. Default is 10 messages per 5 secs. */ 4624 ratelimit_state_init(&sbi->s_err_ratelimit_state, 5 * HZ, 10); 4625 ratelimit_state_init(&sbi->s_warning_ratelimit_state, 5 * HZ, 10); 4626 ratelimit_state_init(&sbi->s_msg_ratelimit_state, 5 * HZ, 10); 4627 4628 kfree(orig_data); 4629 return 0; 4630 4631 cantfind_ext4: 4632 if (!silent) 4633 ext4_msg(sb, KERN_ERR, "VFS: Can't find ext4 filesystem"); 4634 goto failed_mount; 4635 4636 #ifdef CONFIG_QUOTA 4637 failed_mount8: 4638 ext4_unregister_sysfs(sb); 4639 #endif 4640 failed_mount7: 4641 ext4_unregister_li_request(sb); 4642 failed_mount6: 4643 ext4_mb_release(sb); 4644 if (sbi->s_flex_groups) 4645 kvfree(sbi->s_flex_groups); 4646 percpu_counter_destroy(&sbi->s_freeclusters_counter); 4647 percpu_counter_destroy(&sbi->s_freeinodes_counter); 4648 percpu_counter_destroy(&sbi->s_dirs_counter); 4649 percpu_counter_destroy(&sbi->s_dirtyclusters_counter); 4650 percpu_free_rwsem(&sbi->s_journal_flag_rwsem); 4651 failed_mount5: 4652 ext4_ext_release(sb); 4653 ext4_release_system_zone(sb); 4654 failed_mount4a: 4655 dput(sb->s_root); 4656 sb->s_root = NULL; 4657 failed_mount4: 4658 ext4_msg(sb, KERN_ERR, "mount failed"); 4659 if (EXT4_SB(sb)->rsv_conversion_wq) 4660 destroy_workqueue(EXT4_SB(sb)->rsv_conversion_wq); 4661 failed_mount_wq: 4662 ext4_xattr_destroy_cache(sbi->s_ea_inode_cache); 4663 sbi->s_ea_inode_cache = NULL; 4664 4665 ext4_xattr_destroy_cache(sbi->s_ea_block_cache); 4666 sbi->s_ea_block_cache = NULL; 4667 4668 if (sbi->s_journal) { 4669 jbd2_journal_destroy(sbi->s_journal); 4670 sbi->s_journal = NULL; 4671 } 4672 failed_mount3a: 4673 ext4_es_unregister_shrinker(sbi); 4674 failed_mount3: 4675 del_timer_sync(&sbi->s_err_report); 4676 if (sbi->s_mmp_tsk) 4677 kthread_stop(sbi->s_mmp_tsk); 4678 failed_mount2: 4679 for (i = 0; i < db_count; i++) 4680 brelse(sbi->s_group_desc[i]); 4681 kvfree(sbi->s_group_desc); 4682 failed_mount: 4683 if (sbi->s_chksum_driver) 4684 crypto_free_shash(sbi->s_chksum_driver); 4685 4686 #ifdef CONFIG_UNICODE 4687 utf8_unload(sbi->s_encoding); 4688 #endif 4689 4690 #ifdef CONFIG_QUOTA 4691 for (i = 0; i < EXT4_MAXQUOTAS; i++) 4692 kfree(get_qf_name(sb, sbi, i)); 4693 #endif 4694 ext4_blkdev_remove(sbi); 4695 brelse(bh); 4696 out_fail: 4697 sb->s_fs_info = NULL; 4698 kfree(sbi->s_blockgroup_lock); 4699 out_free_base: 4700 kfree(sbi); 4701 kfree(orig_data); 4702 fs_put_dax(dax_dev); 4703 return err ? err : ret; 4704 } 4705 4706 /* 4707 * Setup any per-fs journal parameters now. We'll do this both on 4708 * initial mount, once the journal has been initialised but before we've 4709 * done any recovery; and again on any subsequent remount. 4710 */ 4711 static void ext4_init_journal_params(struct super_block *sb, journal_t *journal) 4712 { 4713 struct ext4_sb_info *sbi = EXT4_SB(sb); 4714 4715 journal->j_commit_interval = sbi->s_commit_interval; 4716 journal->j_min_batch_time = sbi->s_min_batch_time; 4717 journal->j_max_batch_time = sbi->s_max_batch_time; 4718 4719 write_lock(&journal->j_state_lock); 4720 if (test_opt(sb, BARRIER)) 4721 journal->j_flags |= JBD2_BARRIER; 4722 else 4723 journal->j_flags &= ~JBD2_BARRIER; 4724 if (test_opt(sb, DATA_ERR_ABORT)) 4725 journal->j_flags |= JBD2_ABORT_ON_SYNCDATA_ERR; 4726 else 4727 journal->j_flags &= ~JBD2_ABORT_ON_SYNCDATA_ERR; 4728 write_unlock(&journal->j_state_lock); 4729 } 4730 4731 static struct inode *ext4_get_journal_inode(struct super_block *sb, 4732 unsigned int journal_inum) 4733 { 4734 struct inode *journal_inode; 4735 4736 /* 4737 * Test for the existence of a valid inode on disk. Bad things 4738 * happen if we iget() an unused inode, as the subsequent iput() 4739 * will try to delete it. 4740 */ 4741 journal_inode = ext4_iget(sb, journal_inum, EXT4_IGET_SPECIAL); 4742 if (IS_ERR(journal_inode)) { 4743 ext4_msg(sb, KERN_ERR, "no journal found"); 4744 return NULL; 4745 } 4746 if (!journal_inode->i_nlink) { 4747 make_bad_inode(journal_inode); 4748 iput(journal_inode); 4749 ext4_msg(sb, KERN_ERR, "journal inode is deleted"); 4750 return NULL; 4751 } 4752 4753 jbd_debug(2, "Journal inode found at %p: %lld bytes\n", 4754 journal_inode, journal_inode->i_size); 4755 if (!S_ISREG(journal_inode->i_mode)) { 4756 ext4_msg(sb, KERN_ERR, "invalid journal inode"); 4757 iput(journal_inode); 4758 return NULL; 4759 } 4760 return journal_inode; 4761 } 4762 4763 static journal_t *ext4_get_journal(struct super_block *sb, 4764 unsigned int journal_inum) 4765 { 4766 struct inode *journal_inode; 4767 journal_t *journal; 4768 4769 BUG_ON(!ext4_has_feature_journal(sb)); 4770 4771 journal_inode = ext4_get_journal_inode(sb, journal_inum); 4772 if (!journal_inode) 4773 return NULL; 4774 4775 journal = jbd2_journal_init_inode(journal_inode); 4776 if (!journal) { 4777 ext4_msg(sb, KERN_ERR, "Could not load journal inode"); 4778 iput(journal_inode); 4779 return NULL; 4780 } 4781 journal->j_private = sb; 4782 ext4_init_journal_params(sb, journal); 4783 return journal; 4784 } 4785 4786 static journal_t *ext4_get_dev_journal(struct super_block *sb, 4787 dev_t j_dev) 4788 { 4789 struct buffer_head *bh; 4790 journal_t *journal; 4791 ext4_fsblk_t start; 4792 ext4_fsblk_t len; 4793 int hblock, blocksize; 4794 ext4_fsblk_t sb_block; 4795 unsigned long offset; 4796 struct ext4_super_block *es; 4797 struct block_device *bdev; 4798 4799 BUG_ON(!ext4_has_feature_journal(sb)); 4800 4801 bdev = ext4_blkdev_get(j_dev, sb); 4802 if (bdev == NULL) 4803 return NULL; 4804 4805 blocksize = sb->s_blocksize; 4806 hblock = bdev_logical_block_size(bdev); 4807 if (blocksize < hblock) { 4808 ext4_msg(sb, KERN_ERR, 4809 "blocksize too small for journal device"); 4810 goto out_bdev; 4811 } 4812 4813 sb_block = EXT4_MIN_BLOCK_SIZE / blocksize; 4814 offset = EXT4_MIN_BLOCK_SIZE % blocksize; 4815 set_blocksize(bdev, blocksize); 4816 if (!(bh = __bread(bdev, sb_block, blocksize))) { 4817 ext4_msg(sb, KERN_ERR, "couldn't read superblock of " 4818 "external journal"); 4819 goto out_bdev; 4820 } 4821 4822 es = (struct ext4_super_block *) (bh->b_data + offset); 4823 if ((le16_to_cpu(es->s_magic) != EXT4_SUPER_MAGIC) || 4824 !(le32_to_cpu(es->s_feature_incompat) & 4825 EXT4_FEATURE_INCOMPAT_JOURNAL_DEV)) { 4826 ext4_msg(sb, KERN_ERR, "external journal has " 4827 "bad superblock"); 4828 brelse(bh); 4829 goto out_bdev; 4830 } 4831 4832 if ((le32_to_cpu(es->s_feature_ro_compat) & 4833 EXT4_FEATURE_RO_COMPAT_METADATA_CSUM) && 4834 es->s_checksum != ext4_superblock_csum(sb, es)) { 4835 ext4_msg(sb, KERN_ERR, "external journal has " 4836 "corrupt superblock"); 4837 brelse(bh); 4838 goto out_bdev; 4839 } 4840 4841 if (memcmp(EXT4_SB(sb)->s_es->s_journal_uuid, es->s_uuid, 16)) { 4842 ext4_msg(sb, KERN_ERR, "journal UUID does not match"); 4843 brelse(bh); 4844 goto out_bdev; 4845 } 4846 4847 len = ext4_blocks_count(es); 4848 start = sb_block + 1; 4849 brelse(bh); /* we're done with the superblock */ 4850 4851 journal = jbd2_journal_init_dev(bdev, sb->s_bdev, 4852 start, len, blocksize); 4853 if (!journal) { 4854 ext4_msg(sb, KERN_ERR, "failed to create device journal"); 4855 goto out_bdev; 4856 } 4857 journal->j_private = sb; 4858 ll_rw_block(REQ_OP_READ, REQ_META | REQ_PRIO, 1, &journal->j_sb_buffer); 4859 wait_on_buffer(journal->j_sb_buffer); 4860 if (!buffer_uptodate(journal->j_sb_buffer)) { 4861 ext4_msg(sb, KERN_ERR, "I/O error on journal device"); 4862 goto out_journal; 4863 } 4864 if (be32_to_cpu(journal->j_superblock->s_nr_users) != 1) { 4865 ext4_msg(sb, KERN_ERR, "External journal has more than one " 4866 "user (unsupported) - %d", 4867 be32_to_cpu(journal->j_superblock->s_nr_users)); 4868 goto out_journal; 4869 } 4870 EXT4_SB(sb)->journal_bdev = bdev; 4871 ext4_init_journal_params(sb, journal); 4872 return journal; 4873 4874 out_journal: 4875 jbd2_journal_destroy(journal); 4876 out_bdev: 4877 ext4_blkdev_put(bdev); 4878 return NULL; 4879 } 4880 4881 static int ext4_load_journal(struct super_block *sb, 4882 struct ext4_super_block *es, 4883 unsigned long journal_devnum) 4884 { 4885 journal_t *journal; 4886 unsigned int journal_inum = le32_to_cpu(es->s_journal_inum); 4887 dev_t journal_dev; 4888 int err = 0; 4889 int really_read_only; 4890 4891 BUG_ON(!ext4_has_feature_journal(sb)); 4892 4893 if (journal_devnum && 4894 journal_devnum != le32_to_cpu(es->s_journal_dev)) { 4895 ext4_msg(sb, KERN_INFO, "external journal device major/minor " 4896 "numbers have changed"); 4897 journal_dev = new_decode_dev(journal_devnum); 4898 } else 4899 journal_dev = new_decode_dev(le32_to_cpu(es->s_journal_dev)); 4900 4901 really_read_only = bdev_read_only(sb->s_bdev); 4902 4903 /* 4904 * Are we loading a blank journal or performing recovery after a 4905 * crash? For recovery, we need to check in advance whether we 4906 * can get read-write access to the device. 4907 */ 4908 if (ext4_has_feature_journal_needs_recovery(sb)) { 4909 if (sb_rdonly(sb)) { 4910 ext4_msg(sb, KERN_INFO, "INFO: recovery " 4911 "required on readonly filesystem"); 4912 if (really_read_only) { 4913 ext4_msg(sb, KERN_ERR, "write access " 4914 "unavailable, cannot proceed " 4915 "(try mounting with noload)"); 4916 return -EROFS; 4917 } 4918 ext4_msg(sb, KERN_INFO, "write access will " 4919 "be enabled during recovery"); 4920 } 4921 } 4922 4923 if (journal_inum && journal_dev) { 4924 ext4_msg(sb, KERN_ERR, "filesystem has both journal " 4925 "and inode journals!"); 4926 return -EINVAL; 4927 } 4928 4929 if (journal_inum) { 4930 if (!(journal = ext4_get_journal(sb, journal_inum))) 4931 return -EINVAL; 4932 } else { 4933 if (!(journal = ext4_get_dev_journal(sb, journal_dev))) 4934 return -EINVAL; 4935 } 4936 4937 if (!(journal->j_flags & JBD2_BARRIER)) 4938 ext4_msg(sb, KERN_INFO, "barriers disabled"); 4939 4940 if (!ext4_has_feature_journal_needs_recovery(sb)) 4941 err = jbd2_journal_wipe(journal, !really_read_only); 4942 if (!err) { 4943 char *save = kmalloc(EXT4_S_ERR_LEN, GFP_KERNEL); 4944 if (save) 4945 memcpy(save, ((char *) es) + 4946 EXT4_S_ERR_START, EXT4_S_ERR_LEN); 4947 err = jbd2_journal_load(journal); 4948 if (save) 4949 memcpy(((char *) es) + EXT4_S_ERR_START, 4950 save, EXT4_S_ERR_LEN); 4951 kfree(save); 4952 } 4953 4954 if (err) { 4955 ext4_msg(sb, KERN_ERR, "error loading journal"); 4956 jbd2_journal_destroy(journal); 4957 return err; 4958 } 4959 4960 EXT4_SB(sb)->s_journal = journal; 4961 ext4_clear_journal_err(sb, es); 4962 4963 if (!really_read_only && journal_devnum && 4964 journal_devnum != le32_to_cpu(es->s_journal_dev)) { 4965 es->s_journal_dev = cpu_to_le32(journal_devnum); 4966 4967 /* Make sure we flush the recovery flag to disk. */ 4968 ext4_commit_super(sb, 1); 4969 } 4970 4971 return 0; 4972 } 4973 4974 static int ext4_commit_super(struct super_block *sb, int sync) 4975 { 4976 struct ext4_super_block *es = EXT4_SB(sb)->s_es; 4977 struct buffer_head *sbh = EXT4_SB(sb)->s_sbh; 4978 int error = 0; 4979 4980 if (!sbh || block_device_ejected(sb)) 4981 return error; 4982 4983 /* 4984 * The superblock bh should be mapped, but it might not be if the 4985 * device was hot-removed. Not much we can do but fail the I/O. 4986 */ 4987 if (!buffer_mapped(sbh)) 4988 return error; 4989 4990 /* 4991 * If the file system is mounted read-only, don't update the 4992 * superblock write time. This avoids updating the superblock 4993 * write time when we are mounting the root file system 4994 * read/only but we need to replay the journal; at that point, 4995 * for people who are east of GMT and who make their clock 4996 * tick in localtime for Windows bug-for-bug compatibility, 4997 * the clock is set in the future, and this will cause e2fsck 4998 * to complain and force a full file system check. 4999 */ 5000 if (!(sb->s_flags & SB_RDONLY)) 5001 ext4_update_tstamp(es, s_wtime); 5002 if (sb->s_bdev->bd_part) 5003 es->s_kbytes_written = 5004 cpu_to_le64(EXT4_SB(sb)->s_kbytes_written + 5005 ((part_stat_read(sb->s_bdev->bd_part, 5006 sectors[STAT_WRITE]) - 5007 EXT4_SB(sb)->s_sectors_written_start) >> 1)); 5008 else 5009 es->s_kbytes_written = 5010 cpu_to_le64(EXT4_SB(sb)->s_kbytes_written); 5011 if (percpu_counter_initialized(&EXT4_SB(sb)->s_freeclusters_counter)) 5012 ext4_free_blocks_count_set(es, 5013 EXT4_C2B(EXT4_SB(sb), percpu_counter_sum_positive( 5014 &EXT4_SB(sb)->s_freeclusters_counter))); 5015 if (percpu_counter_initialized(&EXT4_SB(sb)->s_freeinodes_counter)) 5016 es->s_free_inodes_count = 5017 cpu_to_le32(percpu_counter_sum_positive( 5018 &EXT4_SB(sb)->s_freeinodes_counter)); 5019 BUFFER_TRACE(sbh, "marking dirty"); 5020 ext4_superblock_csum_set(sb); 5021 if (sync) 5022 lock_buffer(sbh); 5023 if (buffer_write_io_error(sbh) || !buffer_uptodate(sbh)) { 5024 /* 5025 * Oh, dear. A previous attempt to write the 5026 * superblock failed. This could happen because the 5027 * USB device was yanked out. Or it could happen to 5028 * be a transient write error and maybe the block will 5029 * be remapped. Nothing we can do but to retry the 5030 * write and hope for the best. 5031 */ 5032 ext4_msg(sb, KERN_ERR, "previous I/O error to " 5033 "superblock detected"); 5034 clear_buffer_write_io_error(sbh); 5035 set_buffer_uptodate(sbh); 5036 } 5037 mark_buffer_dirty(sbh); 5038 if (sync) { 5039 unlock_buffer(sbh); 5040 error = __sync_dirty_buffer(sbh, 5041 REQ_SYNC | (test_opt(sb, BARRIER) ? REQ_FUA : 0)); 5042 if (buffer_write_io_error(sbh)) { 5043 ext4_msg(sb, KERN_ERR, "I/O error while writing " 5044 "superblock"); 5045 clear_buffer_write_io_error(sbh); 5046 set_buffer_uptodate(sbh); 5047 } 5048 } 5049 return error; 5050 } 5051 5052 /* 5053 * Have we just finished recovery? If so, and if we are mounting (or 5054 * remounting) the filesystem readonly, then we will end up with a 5055 * consistent fs on disk. Record that fact. 5056 */ 5057 static void ext4_mark_recovery_complete(struct super_block *sb, 5058 struct ext4_super_block *es) 5059 { 5060 journal_t *journal = EXT4_SB(sb)->s_journal; 5061 5062 if (!ext4_has_feature_journal(sb)) { 5063 BUG_ON(journal != NULL); 5064 return; 5065 } 5066 jbd2_journal_lock_updates(journal); 5067 if (jbd2_journal_flush(journal) < 0) 5068 goto out; 5069 5070 if (ext4_has_feature_journal_needs_recovery(sb) && sb_rdonly(sb)) { 5071 ext4_clear_feature_journal_needs_recovery(sb); 5072 ext4_commit_super(sb, 1); 5073 } 5074 5075 out: 5076 jbd2_journal_unlock_updates(journal); 5077 } 5078 5079 /* 5080 * If we are mounting (or read-write remounting) a filesystem whose journal 5081 * has recorded an error from a previous lifetime, move that error to the 5082 * main filesystem now. 5083 */ 5084 static void ext4_clear_journal_err(struct super_block *sb, 5085 struct ext4_super_block *es) 5086 { 5087 journal_t *journal; 5088 int j_errno; 5089 const char *errstr; 5090 5091 BUG_ON(!ext4_has_feature_journal(sb)); 5092 5093 journal = EXT4_SB(sb)->s_journal; 5094 5095 /* 5096 * Now check for any error status which may have been recorded in the 5097 * journal by a prior ext4_error() or ext4_abort() 5098 */ 5099 5100 j_errno = jbd2_journal_errno(journal); 5101 if (j_errno) { 5102 char nbuf[16]; 5103 5104 errstr = ext4_decode_error(sb, j_errno, nbuf); 5105 ext4_warning(sb, "Filesystem error recorded " 5106 "from previous mount: %s", errstr); 5107 ext4_warning(sb, "Marking fs in need of filesystem check."); 5108 5109 EXT4_SB(sb)->s_mount_state |= EXT4_ERROR_FS; 5110 es->s_state |= cpu_to_le16(EXT4_ERROR_FS); 5111 ext4_commit_super(sb, 1); 5112 5113 jbd2_journal_clear_err(journal); 5114 jbd2_journal_update_sb_errno(journal); 5115 } 5116 } 5117 5118 /* 5119 * Force the running and committing transactions to commit, 5120 * and wait on the commit. 5121 */ 5122 int ext4_force_commit(struct super_block *sb) 5123 { 5124 journal_t *journal; 5125 5126 if (sb_rdonly(sb)) 5127 return 0; 5128 5129 journal = EXT4_SB(sb)->s_journal; 5130 return ext4_journal_force_commit(journal); 5131 } 5132 5133 static int ext4_sync_fs(struct super_block *sb, int wait) 5134 { 5135 int ret = 0; 5136 tid_t target; 5137 bool needs_barrier = false; 5138 struct ext4_sb_info *sbi = EXT4_SB(sb); 5139 5140 if (unlikely(ext4_forced_shutdown(sbi))) 5141 return 0; 5142 5143 trace_ext4_sync_fs(sb, wait); 5144 flush_workqueue(sbi->rsv_conversion_wq); 5145 /* 5146 * Writeback quota in non-journalled quota case - journalled quota has 5147 * no dirty dquots 5148 */ 5149 dquot_writeback_dquots(sb, -1); 5150 /* 5151 * Data writeback is possible w/o journal transaction, so barrier must 5152 * being sent at the end of the function. But we can skip it if 5153 * transaction_commit will do it for us. 5154 */ 5155 if (sbi->s_journal) { 5156 target = jbd2_get_latest_transaction(sbi->s_journal); 5157 if (wait && sbi->s_journal->j_flags & JBD2_BARRIER && 5158 !jbd2_trans_will_send_data_barrier(sbi->s_journal, target)) 5159 needs_barrier = true; 5160 5161 if (jbd2_journal_start_commit(sbi->s_journal, &target)) { 5162 if (wait) 5163 ret = jbd2_log_wait_commit(sbi->s_journal, 5164 target); 5165 } 5166 } else if (wait && test_opt(sb, BARRIER)) 5167 needs_barrier = true; 5168 if (needs_barrier) { 5169 int err; 5170 err = blkdev_issue_flush(sb->s_bdev, GFP_KERNEL, NULL); 5171 if (!ret) 5172 ret = err; 5173 } 5174 5175 return ret; 5176 } 5177 5178 /* 5179 * LVM calls this function before a (read-only) snapshot is created. This 5180 * gives us a chance to flush the journal completely and mark the fs clean. 5181 * 5182 * Note that only this function cannot bring a filesystem to be in a clean 5183 * state independently. It relies on upper layer to stop all data & metadata 5184 * modifications. 5185 */ 5186 static int ext4_freeze(struct super_block *sb) 5187 { 5188 int error = 0; 5189 journal_t *journal; 5190 5191 if (sb_rdonly(sb)) 5192 return 0; 5193 5194 journal = EXT4_SB(sb)->s_journal; 5195 5196 if (journal) { 5197 /* Now we set up the journal barrier. */ 5198 jbd2_journal_lock_updates(journal); 5199 5200 /* 5201 * Don't clear the needs_recovery flag if we failed to 5202 * flush the journal. 5203 */ 5204 error = jbd2_journal_flush(journal); 5205 if (error < 0) 5206 goto out; 5207 5208 /* Journal blocked and flushed, clear needs_recovery flag. */ 5209 ext4_clear_feature_journal_needs_recovery(sb); 5210 } 5211 5212 error = ext4_commit_super(sb, 1); 5213 out: 5214 if (journal) 5215 /* we rely on upper layer to stop further updates */ 5216 jbd2_journal_unlock_updates(journal); 5217 return error; 5218 } 5219 5220 /* 5221 * Called by LVM after the snapshot is done. We need to reset the RECOVER 5222 * flag here, even though the filesystem is not technically dirty yet. 5223 */ 5224 static int ext4_unfreeze(struct super_block *sb) 5225 { 5226 if (sb_rdonly(sb) || ext4_forced_shutdown(EXT4_SB(sb))) 5227 return 0; 5228 5229 if (EXT4_SB(sb)->s_journal) { 5230 /* Reset the needs_recovery flag before the fs is unlocked. */ 5231 ext4_set_feature_journal_needs_recovery(sb); 5232 } 5233 5234 ext4_commit_super(sb, 1); 5235 return 0; 5236 } 5237 5238 /* 5239 * Structure to save mount options for ext4_remount's benefit 5240 */ 5241 struct ext4_mount_options { 5242 unsigned long s_mount_opt; 5243 unsigned long s_mount_opt2; 5244 kuid_t s_resuid; 5245 kgid_t s_resgid; 5246 unsigned long s_commit_interval; 5247 u32 s_min_batch_time, s_max_batch_time; 5248 #ifdef CONFIG_QUOTA 5249 int s_jquota_fmt; 5250 char *s_qf_names[EXT4_MAXQUOTAS]; 5251 #endif 5252 }; 5253 5254 static int ext4_remount(struct super_block *sb, int *flags, char *data) 5255 { 5256 struct ext4_super_block *es; 5257 struct ext4_sb_info *sbi = EXT4_SB(sb); 5258 unsigned long old_sb_flags; 5259 struct ext4_mount_options old_opts; 5260 int enable_quota = 0; 5261 ext4_group_t g; 5262 unsigned int journal_ioprio = DEFAULT_JOURNAL_IOPRIO; 5263 int err = 0; 5264 #ifdef CONFIG_QUOTA 5265 int i, j; 5266 char *to_free[EXT4_MAXQUOTAS]; 5267 #endif 5268 char *orig_data = kstrdup(data, GFP_KERNEL); 5269 5270 if (data && !orig_data) 5271 return -ENOMEM; 5272 5273 /* Store the original options */ 5274 old_sb_flags = sb->s_flags; 5275 old_opts.s_mount_opt = sbi->s_mount_opt; 5276 old_opts.s_mount_opt2 = sbi->s_mount_opt2; 5277 old_opts.s_resuid = sbi->s_resuid; 5278 old_opts.s_resgid = sbi->s_resgid; 5279 old_opts.s_commit_interval = sbi->s_commit_interval; 5280 old_opts.s_min_batch_time = sbi->s_min_batch_time; 5281 old_opts.s_max_batch_time = sbi->s_max_batch_time; 5282 #ifdef CONFIG_QUOTA 5283 old_opts.s_jquota_fmt = sbi->s_jquota_fmt; 5284 for (i = 0; i < EXT4_MAXQUOTAS; i++) 5285 if (sbi->s_qf_names[i]) { 5286 char *qf_name = get_qf_name(sb, sbi, i); 5287 5288 old_opts.s_qf_names[i] = kstrdup(qf_name, GFP_KERNEL); 5289 if (!old_opts.s_qf_names[i]) { 5290 for (j = 0; j < i; j++) 5291 kfree(old_opts.s_qf_names[j]); 5292 kfree(orig_data); 5293 return -ENOMEM; 5294 } 5295 } else 5296 old_opts.s_qf_names[i] = NULL; 5297 #endif 5298 if (sbi->s_journal && sbi->s_journal->j_task->io_context) 5299 journal_ioprio = sbi->s_journal->j_task->io_context->ioprio; 5300 5301 if (!parse_options(data, sb, NULL, &journal_ioprio, 1)) { 5302 err = -EINVAL; 5303 goto restore_opts; 5304 } 5305 5306 if ((old_opts.s_mount_opt & EXT4_MOUNT_JOURNAL_CHECKSUM) ^ 5307 test_opt(sb, JOURNAL_CHECKSUM)) { 5308 ext4_msg(sb, KERN_ERR, "changing journal_checksum " 5309 "during remount not supported; ignoring"); 5310 sbi->s_mount_opt ^= EXT4_MOUNT_JOURNAL_CHECKSUM; 5311 } 5312 5313 if (test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_JOURNAL_DATA) { 5314 if (test_opt2(sb, EXPLICIT_DELALLOC)) { 5315 ext4_msg(sb, KERN_ERR, "can't mount with " 5316 "both data=journal and delalloc"); 5317 err = -EINVAL; 5318 goto restore_opts; 5319 } 5320 if (test_opt(sb, DIOREAD_NOLOCK)) { 5321 ext4_msg(sb, KERN_ERR, "can't mount with " 5322 "both data=journal and dioread_nolock"); 5323 err = -EINVAL; 5324 goto restore_opts; 5325 } 5326 if (test_opt(sb, DAX)) { 5327 ext4_msg(sb, KERN_ERR, "can't mount with " 5328 "both data=journal and dax"); 5329 err = -EINVAL; 5330 goto restore_opts; 5331 } 5332 } else if (test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_ORDERED_DATA) { 5333 if (test_opt(sb, JOURNAL_ASYNC_COMMIT)) { 5334 ext4_msg(sb, KERN_ERR, "can't mount with " 5335 "journal_async_commit in data=ordered mode"); 5336 err = -EINVAL; 5337 goto restore_opts; 5338 } 5339 } 5340 5341 if ((sbi->s_mount_opt ^ old_opts.s_mount_opt) & EXT4_MOUNT_NO_MBCACHE) { 5342 ext4_msg(sb, KERN_ERR, "can't enable nombcache during remount"); 5343 err = -EINVAL; 5344 goto restore_opts; 5345 } 5346 5347 if ((sbi->s_mount_opt ^ old_opts.s_mount_opt) & EXT4_MOUNT_DAX) { 5348 ext4_msg(sb, KERN_WARNING, "warning: refusing change of " 5349 "dax flag with busy inodes while remounting"); 5350 sbi->s_mount_opt ^= EXT4_MOUNT_DAX; 5351 } 5352 5353 if (sbi->s_mount_flags & EXT4_MF_FS_ABORTED) 5354 ext4_abort(sb, "Abort forced by user"); 5355 5356 sb->s_flags = (sb->s_flags & ~SB_POSIXACL) | 5357 (test_opt(sb, POSIX_ACL) ? SB_POSIXACL : 0); 5358 5359 es = sbi->s_es; 5360 5361 if (sbi->s_journal) { 5362 ext4_init_journal_params(sb, sbi->s_journal); 5363 set_task_ioprio(sbi->s_journal->j_task, journal_ioprio); 5364 } 5365 5366 if (*flags & SB_LAZYTIME) 5367 sb->s_flags |= SB_LAZYTIME; 5368 5369 if ((bool)(*flags & SB_RDONLY) != sb_rdonly(sb)) { 5370 if (sbi->s_mount_flags & EXT4_MF_FS_ABORTED) { 5371 err = -EROFS; 5372 goto restore_opts; 5373 } 5374 5375 if (*flags & SB_RDONLY) { 5376 err = sync_filesystem(sb); 5377 if (err < 0) 5378 goto restore_opts; 5379 err = dquot_suspend(sb, -1); 5380 if (err < 0) 5381 goto restore_opts; 5382 5383 /* 5384 * First of all, the unconditional stuff we have to do 5385 * to disable replay of the journal when we next remount 5386 */ 5387 sb->s_flags |= SB_RDONLY; 5388 5389 /* 5390 * OK, test if we are remounting a valid rw partition 5391 * readonly, and if so set the rdonly flag and then 5392 * mark the partition as valid again. 5393 */ 5394 if (!(es->s_state & cpu_to_le16(EXT4_VALID_FS)) && 5395 (sbi->s_mount_state & EXT4_VALID_FS)) 5396 es->s_state = cpu_to_le16(sbi->s_mount_state); 5397 5398 if (sbi->s_journal) 5399 ext4_mark_recovery_complete(sb, es); 5400 if (sbi->s_mmp_tsk) 5401 kthread_stop(sbi->s_mmp_tsk); 5402 } else { 5403 /* Make sure we can mount this feature set readwrite */ 5404 if (ext4_has_feature_readonly(sb) || 5405 !ext4_feature_set_ok(sb, 0)) { 5406 err = -EROFS; 5407 goto restore_opts; 5408 } 5409 /* 5410 * Make sure the group descriptor checksums 5411 * are sane. If they aren't, refuse to remount r/w. 5412 */ 5413 for (g = 0; g < sbi->s_groups_count; g++) { 5414 struct ext4_group_desc *gdp = 5415 ext4_get_group_desc(sb, g, NULL); 5416 5417 if (!ext4_group_desc_csum_verify(sb, g, gdp)) { 5418 ext4_msg(sb, KERN_ERR, 5419 "ext4_remount: Checksum for group %u failed (%u!=%u)", 5420 g, le16_to_cpu(ext4_group_desc_csum(sb, g, gdp)), 5421 le16_to_cpu(gdp->bg_checksum)); 5422 err = -EFSBADCRC; 5423 goto restore_opts; 5424 } 5425 } 5426 5427 /* 5428 * If we have an unprocessed orphan list hanging 5429 * around from a previously readonly bdev mount, 5430 * require a full umount/remount for now. 5431 */ 5432 if (es->s_last_orphan) { 5433 ext4_msg(sb, KERN_WARNING, "Couldn't " 5434 "remount RDWR because of unprocessed " 5435 "orphan inode list. Please " 5436 "umount/remount instead"); 5437 err = -EINVAL; 5438 goto restore_opts; 5439 } 5440 5441 /* 5442 * Mounting a RDONLY partition read-write, so reread 5443 * and store the current valid flag. (It may have 5444 * been changed by e2fsck since we originally mounted 5445 * the partition.) 5446 */ 5447 if (sbi->s_journal) 5448 ext4_clear_journal_err(sb, es); 5449 sbi->s_mount_state = le16_to_cpu(es->s_state); 5450 5451 err = ext4_setup_super(sb, es, 0); 5452 if (err) 5453 goto restore_opts; 5454 5455 sb->s_flags &= ~SB_RDONLY; 5456 if (ext4_has_feature_mmp(sb)) 5457 if (ext4_multi_mount_protect(sb, 5458 le64_to_cpu(es->s_mmp_block))) { 5459 err = -EROFS; 5460 goto restore_opts; 5461 } 5462 enable_quota = 1; 5463 } 5464 } 5465 5466 /* 5467 * Reinitialize lazy itable initialization thread based on 5468 * current settings 5469 */ 5470 if (sb_rdonly(sb) || !test_opt(sb, INIT_INODE_TABLE)) 5471 ext4_unregister_li_request(sb); 5472 else { 5473 ext4_group_t first_not_zeroed; 5474 first_not_zeroed = ext4_has_uninit_itable(sb); 5475 ext4_register_li_request(sb, first_not_zeroed); 5476 } 5477 5478 ext4_setup_system_zone(sb); 5479 if (sbi->s_journal == NULL && !(old_sb_flags & SB_RDONLY)) { 5480 err = ext4_commit_super(sb, 1); 5481 if (err) 5482 goto restore_opts; 5483 } 5484 5485 #ifdef CONFIG_QUOTA 5486 /* Release old quota file names */ 5487 for (i = 0; i < EXT4_MAXQUOTAS; i++) 5488 kfree(old_opts.s_qf_names[i]); 5489 if (enable_quota) { 5490 if (sb_any_quota_suspended(sb)) 5491 dquot_resume(sb, -1); 5492 else if (ext4_has_feature_quota(sb)) { 5493 err = ext4_enable_quotas(sb); 5494 if (err) 5495 goto restore_opts; 5496 } 5497 } 5498 #endif 5499 5500 *flags = (*flags & ~SB_LAZYTIME) | (sb->s_flags & SB_LAZYTIME); 5501 ext4_msg(sb, KERN_INFO, "re-mounted. Opts: %s", orig_data); 5502 kfree(orig_data); 5503 return 0; 5504 5505 restore_opts: 5506 sb->s_flags = old_sb_flags; 5507 sbi->s_mount_opt = old_opts.s_mount_opt; 5508 sbi->s_mount_opt2 = old_opts.s_mount_opt2; 5509 sbi->s_resuid = old_opts.s_resuid; 5510 sbi->s_resgid = old_opts.s_resgid; 5511 sbi->s_commit_interval = old_opts.s_commit_interval; 5512 sbi->s_min_batch_time = old_opts.s_min_batch_time; 5513 sbi->s_max_batch_time = old_opts.s_max_batch_time; 5514 #ifdef CONFIG_QUOTA 5515 sbi->s_jquota_fmt = old_opts.s_jquota_fmt; 5516 for (i = 0; i < EXT4_MAXQUOTAS; i++) { 5517 to_free[i] = get_qf_name(sb, sbi, i); 5518 rcu_assign_pointer(sbi->s_qf_names[i], old_opts.s_qf_names[i]); 5519 } 5520 synchronize_rcu(); 5521 for (i = 0; i < EXT4_MAXQUOTAS; i++) 5522 kfree(to_free[i]); 5523 #endif 5524 kfree(orig_data); 5525 return err; 5526 } 5527 5528 #ifdef CONFIG_QUOTA 5529 static int ext4_statfs_project(struct super_block *sb, 5530 kprojid_t projid, struct kstatfs *buf) 5531 { 5532 struct kqid qid; 5533 struct dquot *dquot; 5534 u64 limit; 5535 u64 curblock; 5536 5537 qid = make_kqid_projid(projid); 5538 dquot = dqget(sb, qid); 5539 if (IS_ERR(dquot)) 5540 return PTR_ERR(dquot); 5541 spin_lock(&dquot->dq_dqb_lock); 5542 5543 limit = (dquot->dq_dqb.dqb_bsoftlimit ? 5544 dquot->dq_dqb.dqb_bsoftlimit : 5545 dquot->dq_dqb.dqb_bhardlimit) >> sb->s_blocksize_bits; 5546 if (limit && buf->f_blocks > limit) { 5547 curblock = (dquot->dq_dqb.dqb_curspace + 5548 dquot->dq_dqb.dqb_rsvspace) >> sb->s_blocksize_bits; 5549 buf->f_blocks = limit; 5550 buf->f_bfree = buf->f_bavail = 5551 (buf->f_blocks > curblock) ? 5552 (buf->f_blocks - curblock) : 0; 5553 } 5554 5555 limit = dquot->dq_dqb.dqb_isoftlimit ? 5556 dquot->dq_dqb.dqb_isoftlimit : 5557 dquot->dq_dqb.dqb_ihardlimit; 5558 if (limit && buf->f_files > limit) { 5559 buf->f_files = limit; 5560 buf->f_ffree = 5561 (buf->f_files > dquot->dq_dqb.dqb_curinodes) ? 5562 (buf->f_files - dquot->dq_dqb.dqb_curinodes) : 0; 5563 } 5564 5565 spin_unlock(&dquot->dq_dqb_lock); 5566 dqput(dquot); 5567 return 0; 5568 } 5569 #endif 5570 5571 static int ext4_statfs(struct dentry *dentry, struct kstatfs *buf) 5572 { 5573 struct super_block *sb = dentry->d_sb; 5574 struct ext4_sb_info *sbi = EXT4_SB(sb); 5575 struct ext4_super_block *es = sbi->s_es; 5576 ext4_fsblk_t overhead = 0, resv_blocks; 5577 u64 fsid; 5578 s64 bfree; 5579 resv_blocks = EXT4_C2B(sbi, atomic64_read(&sbi->s_resv_clusters)); 5580 5581 if (!test_opt(sb, MINIX_DF)) 5582 overhead = sbi->s_overhead; 5583 5584 buf->f_type = EXT4_SUPER_MAGIC; 5585 buf->f_bsize = sb->s_blocksize; 5586 buf->f_blocks = ext4_blocks_count(es) - EXT4_C2B(sbi, overhead); 5587 bfree = percpu_counter_sum_positive(&sbi->s_freeclusters_counter) - 5588 percpu_counter_sum_positive(&sbi->s_dirtyclusters_counter); 5589 /* prevent underflow in case that few free space is available */ 5590 buf->f_bfree = EXT4_C2B(sbi, max_t(s64, bfree, 0)); 5591 buf->f_bavail = buf->f_bfree - 5592 (ext4_r_blocks_count(es) + resv_blocks); 5593 if (buf->f_bfree < (ext4_r_blocks_count(es) + resv_blocks)) 5594 buf->f_bavail = 0; 5595 buf->f_files = le32_to_cpu(es->s_inodes_count); 5596 buf->f_ffree = percpu_counter_sum_positive(&sbi->s_freeinodes_counter); 5597 buf->f_namelen = EXT4_NAME_LEN; 5598 fsid = le64_to_cpup((void *)es->s_uuid) ^ 5599 le64_to_cpup((void *)es->s_uuid + sizeof(u64)); 5600 buf->f_fsid.val[0] = fsid & 0xFFFFFFFFUL; 5601 buf->f_fsid.val[1] = (fsid >> 32) & 0xFFFFFFFFUL; 5602 5603 #ifdef CONFIG_QUOTA 5604 if (ext4_test_inode_flag(dentry->d_inode, EXT4_INODE_PROJINHERIT) && 5605 sb_has_quota_limits_enabled(sb, PRJQUOTA)) 5606 ext4_statfs_project(sb, EXT4_I(dentry->d_inode)->i_projid, buf); 5607 #endif 5608 return 0; 5609 } 5610 5611 5612 #ifdef CONFIG_QUOTA 5613 5614 /* 5615 * Helper functions so that transaction is started before we acquire dqio_sem 5616 * to keep correct lock ordering of transaction > dqio_sem 5617 */ 5618 static inline struct inode *dquot_to_inode(struct dquot *dquot) 5619 { 5620 return sb_dqopt(dquot->dq_sb)->files[dquot->dq_id.type]; 5621 } 5622 5623 static int ext4_write_dquot(struct dquot *dquot) 5624 { 5625 int ret, err; 5626 handle_t *handle; 5627 struct inode *inode; 5628 5629 inode = dquot_to_inode(dquot); 5630 handle = ext4_journal_start(inode, EXT4_HT_QUOTA, 5631 EXT4_QUOTA_TRANS_BLOCKS(dquot->dq_sb)); 5632 if (IS_ERR(handle)) 5633 return PTR_ERR(handle); 5634 ret = dquot_commit(dquot); 5635 err = ext4_journal_stop(handle); 5636 if (!ret) 5637 ret = err; 5638 return ret; 5639 } 5640 5641 static int ext4_acquire_dquot(struct dquot *dquot) 5642 { 5643 int ret, err; 5644 handle_t *handle; 5645 5646 handle = ext4_journal_start(dquot_to_inode(dquot), EXT4_HT_QUOTA, 5647 EXT4_QUOTA_INIT_BLOCKS(dquot->dq_sb)); 5648 if (IS_ERR(handle)) 5649 return PTR_ERR(handle); 5650 ret = dquot_acquire(dquot); 5651 err = ext4_journal_stop(handle); 5652 if (!ret) 5653 ret = err; 5654 return ret; 5655 } 5656 5657 static int ext4_release_dquot(struct dquot *dquot) 5658 { 5659 int ret, err; 5660 handle_t *handle; 5661 5662 handle = ext4_journal_start(dquot_to_inode(dquot), EXT4_HT_QUOTA, 5663 EXT4_QUOTA_DEL_BLOCKS(dquot->dq_sb)); 5664 if (IS_ERR(handle)) { 5665 /* Release dquot anyway to avoid endless cycle in dqput() */ 5666 dquot_release(dquot); 5667 return PTR_ERR(handle); 5668 } 5669 ret = dquot_release(dquot); 5670 err = ext4_journal_stop(handle); 5671 if (!ret) 5672 ret = err; 5673 return ret; 5674 } 5675 5676 static int ext4_mark_dquot_dirty(struct dquot *dquot) 5677 { 5678 struct super_block *sb = dquot->dq_sb; 5679 struct ext4_sb_info *sbi = EXT4_SB(sb); 5680 5681 /* Are we journaling quotas? */ 5682 if (ext4_has_feature_quota(sb) || 5683 sbi->s_qf_names[USRQUOTA] || sbi->s_qf_names[GRPQUOTA]) { 5684 dquot_mark_dquot_dirty(dquot); 5685 return ext4_write_dquot(dquot); 5686 } else { 5687 return dquot_mark_dquot_dirty(dquot); 5688 } 5689 } 5690 5691 static int ext4_write_info(struct super_block *sb, int type) 5692 { 5693 int ret, err; 5694 handle_t *handle; 5695 5696 /* Data block + inode block */ 5697 handle = ext4_journal_start(d_inode(sb->s_root), EXT4_HT_QUOTA, 2); 5698 if (IS_ERR(handle)) 5699 return PTR_ERR(handle); 5700 ret = dquot_commit_info(sb, type); 5701 err = ext4_journal_stop(handle); 5702 if (!ret) 5703 ret = err; 5704 return ret; 5705 } 5706 5707 /* 5708 * Turn on quotas during mount time - we need to find 5709 * the quota file and such... 5710 */ 5711 static int ext4_quota_on_mount(struct super_block *sb, int type) 5712 { 5713 return dquot_quota_on_mount(sb, get_qf_name(sb, EXT4_SB(sb), type), 5714 EXT4_SB(sb)->s_jquota_fmt, type); 5715 } 5716 5717 static void lockdep_set_quota_inode(struct inode *inode, int subclass) 5718 { 5719 struct ext4_inode_info *ei = EXT4_I(inode); 5720 5721 /* The first argument of lockdep_set_subclass has to be 5722 * *exactly* the same as the argument to init_rwsem() --- in 5723 * this case, in init_once() --- or lockdep gets unhappy 5724 * because the name of the lock is set using the 5725 * stringification of the argument to init_rwsem(). 5726 */ 5727 (void) ei; /* shut up clang warning if !CONFIG_LOCKDEP */ 5728 lockdep_set_subclass(&ei->i_data_sem, subclass); 5729 } 5730 5731 /* 5732 * Standard function to be called on quota_on 5733 */ 5734 static int ext4_quota_on(struct super_block *sb, int type, int format_id, 5735 const struct path *path) 5736 { 5737 int err; 5738 5739 if (!test_opt(sb, QUOTA)) 5740 return -EINVAL; 5741 5742 /* Quotafile not on the same filesystem? */ 5743 if (path->dentry->d_sb != sb) 5744 return -EXDEV; 5745 /* Journaling quota? */ 5746 if (EXT4_SB(sb)->s_qf_names[type]) { 5747 /* Quotafile not in fs root? */ 5748 if (path->dentry->d_parent != sb->s_root) 5749 ext4_msg(sb, KERN_WARNING, 5750 "Quota file not on filesystem root. " 5751 "Journaled quota will not work"); 5752 sb_dqopt(sb)->flags |= DQUOT_NOLIST_DIRTY; 5753 } else { 5754 /* 5755 * Clear the flag just in case mount options changed since 5756 * last time. 5757 */ 5758 sb_dqopt(sb)->flags &= ~DQUOT_NOLIST_DIRTY; 5759 } 5760 5761 /* 5762 * When we journal data on quota file, we have to flush journal to see 5763 * all updates to the file when we bypass pagecache... 5764 */ 5765 if (EXT4_SB(sb)->s_journal && 5766 ext4_should_journal_data(d_inode(path->dentry))) { 5767 /* 5768 * We don't need to lock updates but journal_flush() could 5769 * otherwise be livelocked... 5770 */ 5771 jbd2_journal_lock_updates(EXT4_SB(sb)->s_journal); 5772 err = jbd2_journal_flush(EXT4_SB(sb)->s_journal); 5773 jbd2_journal_unlock_updates(EXT4_SB(sb)->s_journal); 5774 if (err) 5775 return err; 5776 } 5777 5778 lockdep_set_quota_inode(path->dentry->d_inode, I_DATA_SEM_QUOTA); 5779 err = dquot_quota_on(sb, type, format_id, path); 5780 if (err) { 5781 lockdep_set_quota_inode(path->dentry->d_inode, 5782 I_DATA_SEM_NORMAL); 5783 } else { 5784 struct inode *inode = d_inode(path->dentry); 5785 handle_t *handle; 5786 5787 /* 5788 * Set inode flags to prevent userspace from messing with quota 5789 * files. If this fails, we return success anyway since quotas 5790 * are already enabled and this is not a hard failure. 5791 */ 5792 inode_lock(inode); 5793 handle = ext4_journal_start(inode, EXT4_HT_QUOTA, 1); 5794 if (IS_ERR(handle)) 5795 goto unlock_inode; 5796 EXT4_I(inode)->i_flags |= EXT4_NOATIME_FL | EXT4_IMMUTABLE_FL; 5797 inode_set_flags(inode, S_NOATIME | S_IMMUTABLE, 5798 S_NOATIME | S_IMMUTABLE); 5799 ext4_mark_inode_dirty(handle, inode); 5800 ext4_journal_stop(handle); 5801 unlock_inode: 5802 inode_unlock(inode); 5803 } 5804 return err; 5805 } 5806 5807 static int ext4_quota_enable(struct super_block *sb, int type, int format_id, 5808 unsigned int flags) 5809 { 5810 int err; 5811 struct inode *qf_inode; 5812 unsigned long qf_inums[EXT4_MAXQUOTAS] = { 5813 le32_to_cpu(EXT4_SB(sb)->s_es->s_usr_quota_inum), 5814 le32_to_cpu(EXT4_SB(sb)->s_es->s_grp_quota_inum), 5815 le32_to_cpu(EXT4_SB(sb)->s_es->s_prj_quota_inum) 5816 }; 5817 5818 BUG_ON(!ext4_has_feature_quota(sb)); 5819 5820 if (!qf_inums[type]) 5821 return -EPERM; 5822 5823 qf_inode = ext4_iget(sb, qf_inums[type], EXT4_IGET_SPECIAL); 5824 if (IS_ERR(qf_inode)) { 5825 ext4_error(sb, "Bad quota inode # %lu", qf_inums[type]); 5826 return PTR_ERR(qf_inode); 5827 } 5828 5829 /* Don't account quota for quota files to avoid recursion */ 5830 qf_inode->i_flags |= S_NOQUOTA; 5831 lockdep_set_quota_inode(qf_inode, I_DATA_SEM_QUOTA); 5832 err = dquot_load_quota_inode(qf_inode, type, format_id, flags); 5833 if (err) 5834 lockdep_set_quota_inode(qf_inode, I_DATA_SEM_NORMAL); 5835 iput(qf_inode); 5836 5837 return err; 5838 } 5839 5840 /* Enable usage tracking for all quota types. */ 5841 static int ext4_enable_quotas(struct super_block *sb) 5842 { 5843 int type, err = 0; 5844 unsigned long qf_inums[EXT4_MAXQUOTAS] = { 5845 le32_to_cpu(EXT4_SB(sb)->s_es->s_usr_quota_inum), 5846 le32_to_cpu(EXT4_SB(sb)->s_es->s_grp_quota_inum), 5847 le32_to_cpu(EXT4_SB(sb)->s_es->s_prj_quota_inum) 5848 }; 5849 bool quota_mopt[EXT4_MAXQUOTAS] = { 5850 test_opt(sb, USRQUOTA), 5851 test_opt(sb, GRPQUOTA), 5852 test_opt(sb, PRJQUOTA), 5853 }; 5854 5855 sb_dqopt(sb)->flags |= DQUOT_QUOTA_SYS_FILE | DQUOT_NOLIST_DIRTY; 5856 for (type = 0; type < EXT4_MAXQUOTAS; type++) { 5857 if (qf_inums[type]) { 5858 err = ext4_quota_enable(sb, type, QFMT_VFS_V1, 5859 DQUOT_USAGE_ENABLED | 5860 (quota_mopt[type] ? DQUOT_LIMITS_ENABLED : 0)); 5861 if (err) { 5862 ext4_warning(sb, 5863 "Failed to enable quota tracking " 5864 "(type=%d, err=%d). Please run " 5865 "e2fsck to fix.", type, err); 5866 for (type--; type >= 0; type--) 5867 dquot_quota_off(sb, type); 5868 5869 return err; 5870 } 5871 } 5872 } 5873 return 0; 5874 } 5875 5876 static int ext4_quota_off(struct super_block *sb, int type) 5877 { 5878 struct inode *inode = sb_dqopt(sb)->files[type]; 5879 handle_t *handle; 5880 int err; 5881 5882 /* Force all delayed allocation blocks to be allocated. 5883 * Caller already holds s_umount sem */ 5884 if (test_opt(sb, DELALLOC)) 5885 sync_filesystem(sb); 5886 5887 if (!inode || !igrab(inode)) 5888 goto out; 5889 5890 err = dquot_quota_off(sb, type); 5891 if (err || ext4_has_feature_quota(sb)) 5892 goto out_put; 5893 5894 inode_lock(inode); 5895 /* 5896 * Update modification times of quota files when userspace can 5897 * start looking at them. If we fail, we return success anyway since 5898 * this is not a hard failure and quotas are already disabled. 5899 */ 5900 handle = ext4_journal_start(inode, EXT4_HT_QUOTA, 1); 5901 if (IS_ERR(handle)) 5902 goto out_unlock; 5903 EXT4_I(inode)->i_flags &= ~(EXT4_NOATIME_FL | EXT4_IMMUTABLE_FL); 5904 inode_set_flags(inode, 0, S_NOATIME | S_IMMUTABLE); 5905 inode->i_mtime = inode->i_ctime = current_time(inode); 5906 ext4_mark_inode_dirty(handle, inode); 5907 ext4_journal_stop(handle); 5908 out_unlock: 5909 inode_unlock(inode); 5910 out_put: 5911 lockdep_set_quota_inode(inode, I_DATA_SEM_NORMAL); 5912 iput(inode); 5913 return err; 5914 out: 5915 return dquot_quota_off(sb, type); 5916 } 5917 5918 /* Read data from quotafile - avoid pagecache and such because we cannot afford 5919 * acquiring the locks... As quota files are never truncated and quota code 5920 * itself serializes the operations (and no one else should touch the files) 5921 * we don't have to be afraid of races */ 5922 static ssize_t ext4_quota_read(struct super_block *sb, int type, char *data, 5923 size_t len, loff_t off) 5924 { 5925 struct inode *inode = sb_dqopt(sb)->files[type]; 5926 ext4_lblk_t blk = off >> EXT4_BLOCK_SIZE_BITS(sb); 5927 int offset = off & (sb->s_blocksize - 1); 5928 int tocopy; 5929 size_t toread; 5930 struct buffer_head *bh; 5931 loff_t i_size = i_size_read(inode); 5932 5933 if (off > i_size) 5934 return 0; 5935 if (off+len > i_size) 5936 len = i_size-off; 5937 toread = len; 5938 while (toread > 0) { 5939 tocopy = sb->s_blocksize - offset < toread ? 5940 sb->s_blocksize - offset : toread; 5941 bh = ext4_bread(NULL, inode, blk, 0); 5942 if (IS_ERR(bh)) 5943 return PTR_ERR(bh); 5944 if (!bh) /* A hole? */ 5945 memset(data, 0, tocopy); 5946 else 5947 memcpy(data, bh->b_data+offset, tocopy); 5948 brelse(bh); 5949 offset = 0; 5950 toread -= tocopy; 5951 data += tocopy; 5952 blk++; 5953 } 5954 return len; 5955 } 5956 5957 /* Write to quotafile (we know the transaction is already started and has 5958 * enough credits) */ 5959 static ssize_t ext4_quota_write(struct super_block *sb, int type, 5960 const char *data, size_t len, loff_t off) 5961 { 5962 struct inode *inode = sb_dqopt(sb)->files[type]; 5963 ext4_lblk_t blk = off >> EXT4_BLOCK_SIZE_BITS(sb); 5964 int err, offset = off & (sb->s_blocksize - 1); 5965 int retries = 0; 5966 struct buffer_head *bh; 5967 handle_t *handle = journal_current_handle(); 5968 5969 if (EXT4_SB(sb)->s_journal && !handle) { 5970 ext4_msg(sb, KERN_WARNING, "Quota write (off=%llu, len=%llu)" 5971 " cancelled because transaction is not started", 5972 (unsigned long long)off, (unsigned long long)len); 5973 return -EIO; 5974 } 5975 /* 5976 * Since we account only one data block in transaction credits, 5977 * then it is impossible to cross a block boundary. 5978 */ 5979 if (sb->s_blocksize - offset < len) { 5980 ext4_msg(sb, KERN_WARNING, "Quota write (off=%llu, len=%llu)" 5981 " cancelled because not block aligned", 5982 (unsigned long long)off, (unsigned long long)len); 5983 return -EIO; 5984 } 5985 5986 do { 5987 bh = ext4_bread(handle, inode, blk, 5988 EXT4_GET_BLOCKS_CREATE | 5989 EXT4_GET_BLOCKS_METADATA_NOFAIL); 5990 } while (IS_ERR(bh) && (PTR_ERR(bh) == -ENOSPC) && 5991 ext4_should_retry_alloc(inode->i_sb, &retries)); 5992 if (IS_ERR(bh)) 5993 return PTR_ERR(bh); 5994 if (!bh) 5995 goto out; 5996 BUFFER_TRACE(bh, "get write access"); 5997 err = ext4_journal_get_write_access(handle, bh); 5998 if (err) { 5999 brelse(bh); 6000 return err; 6001 } 6002 lock_buffer(bh); 6003 memcpy(bh->b_data+offset, data, len); 6004 flush_dcache_page(bh->b_page); 6005 unlock_buffer(bh); 6006 err = ext4_handle_dirty_metadata(handle, NULL, bh); 6007 brelse(bh); 6008 out: 6009 if (inode->i_size < off + len) { 6010 i_size_write(inode, off + len); 6011 EXT4_I(inode)->i_disksize = inode->i_size; 6012 ext4_mark_inode_dirty(handle, inode); 6013 } 6014 return len; 6015 } 6016 #endif 6017 6018 static struct dentry *ext4_mount(struct file_system_type *fs_type, int flags, 6019 const char *dev_name, void *data) 6020 { 6021 return mount_bdev(fs_type, flags, dev_name, data, ext4_fill_super); 6022 } 6023 6024 #if !defined(CONFIG_EXT2_FS) && !defined(CONFIG_EXT2_FS_MODULE) && defined(CONFIG_EXT4_USE_FOR_EXT2) 6025 static inline void register_as_ext2(void) 6026 { 6027 int err = register_filesystem(&ext2_fs_type); 6028 if (err) 6029 printk(KERN_WARNING 6030 "EXT4-fs: Unable to register as ext2 (%d)\n", err); 6031 } 6032 6033 static inline void unregister_as_ext2(void) 6034 { 6035 unregister_filesystem(&ext2_fs_type); 6036 } 6037 6038 static inline int ext2_feature_set_ok(struct super_block *sb) 6039 { 6040 if (ext4_has_unknown_ext2_incompat_features(sb)) 6041 return 0; 6042 if (sb_rdonly(sb)) 6043 return 1; 6044 if (ext4_has_unknown_ext2_ro_compat_features(sb)) 6045 return 0; 6046 return 1; 6047 } 6048 #else 6049 static inline void register_as_ext2(void) { } 6050 static inline void unregister_as_ext2(void) { } 6051 static inline int ext2_feature_set_ok(struct super_block *sb) { return 0; } 6052 #endif 6053 6054 static inline void register_as_ext3(void) 6055 { 6056 int err = register_filesystem(&ext3_fs_type); 6057 if (err) 6058 printk(KERN_WARNING 6059 "EXT4-fs: Unable to register as ext3 (%d)\n", err); 6060 } 6061 6062 static inline void unregister_as_ext3(void) 6063 { 6064 unregister_filesystem(&ext3_fs_type); 6065 } 6066 6067 static inline int ext3_feature_set_ok(struct super_block *sb) 6068 { 6069 if (ext4_has_unknown_ext3_incompat_features(sb)) 6070 return 0; 6071 if (!ext4_has_feature_journal(sb)) 6072 return 0; 6073 if (sb_rdonly(sb)) 6074 return 1; 6075 if (ext4_has_unknown_ext3_ro_compat_features(sb)) 6076 return 0; 6077 return 1; 6078 } 6079 6080 static struct file_system_type ext4_fs_type = { 6081 .owner = THIS_MODULE, 6082 .name = "ext4", 6083 .mount = ext4_mount, 6084 .kill_sb = kill_block_super, 6085 .fs_flags = FS_REQUIRES_DEV, 6086 }; 6087 MODULE_ALIAS_FS("ext4"); 6088 6089 /* Shared across all ext4 file systems */ 6090 wait_queue_head_t ext4__ioend_wq[EXT4_WQ_HASH_SZ]; 6091 6092 static int __init ext4_init_fs(void) 6093 { 6094 int i, err; 6095 6096 ratelimit_state_init(&ext4_mount_msg_ratelimit, 30 * HZ, 64); 6097 ext4_li_info = NULL; 6098 mutex_init(&ext4_li_mtx); 6099 6100 /* Build-time check for flags consistency */ 6101 ext4_check_flag_values(); 6102 6103 for (i = 0; i < EXT4_WQ_HASH_SZ; i++) 6104 init_waitqueue_head(&ext4__ioend_wq[i]); 6105 6106 err = ext4_init_es(); 6107 if (err) 6108 return err; 6109 6110 err = ext4_init_pending(); 6111 if (err) 6112 goto out7; 6113 6114 err = ext4_init_post_read_processing(); 6115 if (err) 6116 goto out6; 6117 6118 err = ext4_init_pageio(); 6119 if (err) 6120 goto out5; 6121 6122 err = ext4_init_system_zone(); 6123 if (err) 6124 goto out4; 6125 6126 err = ext4_init_sysfs(); 6127 if (err) 6128 goto out3; 6129 6130 err = ext4_init_mballoc(); 6131 if (err) 6132 goto out2; 6133 err = init_inodecache(); 6134 if (err) 6135 goto out1; 6136 register_as_ext3(); 6137 register_as_ext2(); 6138 err = register_filesystem(&ext4_fs_type); 6139 if (err) 6140 goto out; 6141 6142 return 0; 6143 out: 6144 unregister_as_ext2(); 6145 unregister_as_ext3(); 6146 destroy_inodecache(); 6147 out1: 6148 ext4_exit_mballoc(); 6149 out2: 6150 ext4_exit_sysfs(); 6151 out3: 6152 ext4_exit_system_zone(); 6153 out4: 6154 ext4_exit_pageio(); 6155 out5: 6156 ext4_exit_post_read_processing(); 6157 out6: 6158 ext4_exit_pending(); 6159 out7: 6160 ext4_exit_es(); 6161 6162 return err; 6163 } 6164 6165 static void __exit ext4_exit_fs(void) 6166 { 6167 ext4_destroy_lazyinit_thread(); 6168 unregister_as_ext2(); 6169 unregister_as_ext3(); 6170 unregister_filesystem(&ext4_fs_type); 6171 destroy_inodecache(); 6172 ext4_exit_mballoc(); 6173 ext4_exit_sysfs(); 6174 ext4_exit_system_zone(); 6175 ext4_exit_pageio(); 6176 ext4_exit_post_read_processing(); 6177 ext4_exit_es(); 6178 ext4_exit_pending(); 6179 } 6180 6181 MODULE_AUTHOR("Remy Card, Stephen Tweedie, Andrew Morton, Andreas Dilger, Theodore Ts'o and others"); 6182 MODULE_DESCRIPTION("Fourth Extended Filesystem"); 6183 MODULE_LICENSE("GPL"); 6184 MODULE_SOFTDEP("pre: crc32c"); 6185 module_init(ext4_init_fs) 6186 module_exit(ext4_exit_fs) 6187